|Publication number||US4370709 A|
|Application number||US 06/174,721|
|Publication date||Jan 25, 1983|
|Filing date||Aug 1, 1980|
|Priority date||Aug 1, 1980|
|Publication number||06174721, 174721, US 4370709 A, US 4370709A, US-A-4370709, US4370709 A, US4370709A|
|Inventors||Robert E. Fosdick|
|Original Assignee||Tracor, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (14), Non-Patent Citations (4), Referenced by (67), Classifications (9), Legal Events (10)|
|External Links: USPTO, USPTO Assignment, Espacenet|
BACKGROUND OF THE INVENTION
SUMMARY OF THE INVENTION
BRIEF DESCRIPTION OF THE DRAWINGS
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
A Block Diagram of the Computer Emulator Architecture
The General Processor Unit
The Register Select Unit
The Microcontroller Unit
The Bus Interface Unit
The Address Selection Unit
The Interrupt Unit
The Op-Code Transform ROM
A Microinstruction Emulation Example
This invention relates to computers. More particularly, the invention relates to a microcoded computer design which executes the instruction set of another computer, otherwise known as a computer emulator.
Every computer has an instruction set which the computer is designed to execute. While the hardware cost to purchase these computers involve an expenditure of money, this initial hardware cost is small in comparison to the cost of producing the computer programs that are run on the computer. For this reason, new applications requiring the use of a computer would best be served economically if the computer selected could execute computer programs already written. Unfortunately, some of these new applications require a computer that is not available. Such requirements as low power consumption, small size, radiation harden circuits, and severe environment operating conditions, along with the ability to execute these pre-written computer programs, make it impossible for existing designs to be acceptable. Usually, the manufacturer of the computer for these already existing usable application programs does not offer a model of computer which meets these requirements.
To solve the problem of supplying a new computer which is able to meet these stringent specification requirements, such as the low power, small size, etc., and at the same time execute the instruction set of an already existing computer design, a new type of computer has evolved--computer emulators. A computer emulator is a special type of computer which uses microcoded instructions in firmware to control its internal circuits in response to the macroinstructions of another computer to achieve the same external results as the macroinstructions achieve in its own computer, and do so without a substantial reduction in program execution time. To the macroinstruction of the other computer, the computer emulator looks the same as its own computer, even though the internal circuit structure of the computer emulator may be entirely different than that for the other computer.
Prior-art computer emulators have used microcoded firmware routines to carry out the instructions of another computer. A mapping ROM is used to map each macroinstruction into one of these microcoded firmware routines that will emulate the macroinstruction. For this design approach, there is much duplication of microcode. This duplication results in an increase in power consumption due to the increase in the dize of the micromemory required to store the microcode. The complexity of the instruction set further controls the size of micromemory, the more complex the instruction set, the larger the memory size and resulting power consumption.
A further problem present in prior-art emulators is the throughput rate of each macroinstruction. The emulator should not significantly reduce the time required to execute each macroinstruction from the normal time for that same macroinstruction to run on its own machine. Because these prior-art emulators executed each macroinstruction, including operand derivation and instruction execution, in a sequential fashion, extra microcycles must be used where the operand deviation phase of the emulation results in more data being required from memory before the execution phase of the instruction can proceed. These extra macrocycles tend to decrease throughput rate of the emulator.
Accordingly, it would be advantageous to provide a low-power, ruggedized computer emulator capable of executing the instruction set of another computer without a significant reduction in macroinstruction emulation time by providing dual microcontrollers operating in parallel to respectively handle the operand derivation phase and the instruction execution phase of each macroinstruction. It would also be advantageous to provide a sectionalized micromemory for storing the microcode routines so that the same microcode used by various macroinstructions need not be duplicated for each macroinstruction, but is available to both microcontrollers on a need to have basis.
In accordance with the present invention, a microcoded central processing unit for emulating the macroinstructions of another computer is disclosed. Each macroinstruction emulation is divided into two phases, the operand address derivation phase and the instruction execution phase. Both phases of the macroinstruction emulation occur simultaneously and in parallel to the extent that it is possible. Where a computer resource is required in both phases at the same time, one phase must pause and wait until that resource become available, and sometimes, one phase must wait until the other phase has completed its operations before it can begin its own operations.
The central processing unit of the present invention includes a micromemory comprising a plurality of micromemory units including at least a right, a middle, and a left micromemory unit. The micromemory has an address space in which the microcoded instructions are stored. Each micromemory unit is separately addressed. Controlling the operand address derivation phase and the instruction execution phase of each macroinstruction emulation is a first and a second microcontroller, respectively.
The first microcontroller is associated with the right micromemory unit and responds to the microcode stored therein for generating the micromemory address of the next microcoded instruction to be outputted from said right micromemory unit. The second microcontroller is associated with the left micromemory unit and responds to the microcode stored therein for generating the micromemory address of the next coded instruction to be outputted from the left micromemory unit. The micromemory address that is generated by either the first or the second microcontroller may be used to address not only its associated micromemory unit but also, each additional micromemory unit in the micromemory. This includes the middle micromemory unit and the micromemory associated with the other microcontroller.
Further included in the present invention is a plurality of address selection units, with each address selection unit associated with one of the plurality of micromemory units comprising the micromemory. Each address selection unit temporarily stores and outputs to its associated micromemory unit a micromemory address from a selected one of the microcontrollers. Each address selection unit generates clocking signals to the CPU resources controlled by the microcode stored in its associated micromemory unit. Each address selection unit receives the micromemory address generated by both the first and the second microcontrollers.
An op-code mapping means is provided for mapping each macroinstruction into the micromemory address space at the starting addresses of the microcoded routines which will control the operand address derivation and the instruction execution phases for each macroinstruction emulation. A bus interface means is provided for coupling the main memory and the peripherals of the computer system to the CPU, and for carrying out selected internal data bus transfers between the various functional blocks that comprise the computer architecture.
Associated with the first and second microcontrollers are first and second processing units, respectively. The processing units respond to the bus interface means and to the microcode to perform logical operation on the digital data of the CPU whereby the emulation of the macroinstructions are achieved. Each processing unit includes an artithmetic logic unit. During the emulation of each macroinstruction, the right and left microcontrollers operate concurrently and in parallel so that the operand address derivation and the execution of each macroinstruction can proceed at the same time. During the emulation of each macroinstruction, selected ones of the address selection units will be inhibited for a period of time from generating their clocking signals when the microcode from a micromemory unit requests CPU resources controlled by another micromemory unit and those resources are not then available for use by the requesting microcode.
The present invention further includes an interrupt unit for handling a priority system of peripheral interrupt requests.
A computer emulator for executing the macroinstructions of another computer constructed in accordance with the preferred embodiment of the invention is illustrated in the accompanying drawings in which,
FIG. 1 is a block diagram of the computer emulator architecture;
FIG. 2 is a more detailed block diagram of the signal connections between the address selection units and their associated micromemory units;
FIG. 3 is a functional block diagram of the general processing unit shown in FIG. 1;
FIG. 4 is a functional block diagram of a typical register select units shown in FIG. 1;
FIGS. 5(a), (b), (c) and (d) comprise a more detailed circuit diagram of the register select unit illustrated in FIG. 4;
FIG. 6 is a format diagram of the instruction set for an emulated computer instruction set;
FIG. 7 is a functional block diagram of the microcontroller shown in FIG. 1;
FIG. 8 is a functional block diagram of the discrete interface illustrated in FIG. 7;
FIG. 9 is a functional block diagram of a typical address selection unit shown in FIG. 1;
FIGS. 10(a) and (b) comprise a more detailed circuit diagram of the address selection unit illustrated in FIG. 9;
FIG. 11 is a functional block diagram of a typical bus innerface unit illustrated in FIG. 1;
FIGS. 12(a), (b), (c), (d) and (e) comprise a more detailed circuit diagram of the bus innerface unit shown in FIG. 11;
FIG. 13 is a functional bock diagram of the bus interface means of the invention as constructed from individual bus interface units shown in FIGS. 11 and 12;
FIG. 14 is a functional block diagram of the interrupt unit shown in FIG. 1; and
FIGS. 15(a), (b) and (c) is a detailed circuit diagram of the interrupt unit as shown in FIG. 14.
Similar reference numerals refer to similar parts throughout the several views of the drawings.
Several of the drawings accompanying the disclosure of the invention are detailed logic circuit diagrams of various functional portions of the invention. In these drawings, where repeated logic circuits occur, rectangular functional blocks have been substituted for those circuits with an appropriate reference number to identify the circuit diagram contained inside each functional block. At least one detailed circuit diagram of the circuits contained in each of these functional blocks are given in at least one of the drawings.
Turning now to the figures, and first to FIG. 1, a functional block diagram of the architecture of the invention is shown. The invention consists of a micromemory 21 consisting of a plurality of micromemory units 20, 22, 24; a plurality of address selection units 26, 28, 30 respectively associated with the left memory unit 20 (LM), the middle memory unit 22 (MM) and the right memory unit 24 (RM) memory units; microcontroller units 1 32 (MCU1) and unit 2 34 (MCU2); and op-code transform ROM 36; interrupt unit 38; input/output peripheral interface unit 40; first and second processing means 41, 43; and bus interface units 58 (BIU4), 60 (BIU3), 62 (BIU2) and 64 (BIU1).
The micromemory unit 21 is used to store the firmware microcoded instructions which are used to control the internal circuits of the processor to emulate the macroinstructions of the other computer. The micromemory 21 consists of three micromemory units, left micromemory 20 (LM), middle micromemory unit 22 (MM) and right micromemory unit 24 (RM). Associated with each micromemory unit 20, 22, 24 is an associated address selection unit (ASU) 26, 28, 30. The output from each ASU is a 10-bit binary address to its associated micromemory unit. Inputted to each ASU are two 10-bit binary addresses, one address each from MCU1 32 and MCU2 34. As illustrated in FIG. 1, the two 10-bit binary addresses supplied from the MCU units are inputted either to the A or the B inputs of the ASUs. For the MCU1 32, the 10-bit binary address is supplied to the A input of ASU1 26 and to the B inputs of ASU2 28 and ASU3 30. For MCU2 34, the 10-bit address is inputted to the A input of the ASU2 28 and ASU3 30 and to the B input of the ASU1 26. The internal circuits of the ASUs are designed to provide a priority to the A input address if a request is issued to the address selection unit to input both the A and the B address at the same time.
Referring now to FIG. 2, a more detailed block diagram of the control signal connections between the three micromemory units 20, 22, 24 and their associated ASU units 26, 28, 30 is shown. The illustrated signal lines do not show the address lines. Each ASU unit has a REQ A and a REQ B input control signal. A REQ A requests the 10-bit address on the A input of the ASU to be outputted to the associated micromemory unit while a REQ B requests the 10-bit binary address on the B input lines be outputted. As previously mentioned, when both the REQ A and REQ B signal occur simultaneously, priority is given to the REQ A input signal and the 10-bit address on the A input lines will be temporarily stored and outputted to the associated micromemory unit. As will be more fully discussed below, MCU1 normally addresses the LM 20 for the microcode stored therein that is needed during the execution phase of each macroinstruction emulation sequence, while the MCU2 addresses the RM 24 to obtain the microcode to control the operand derivation phase of the macroinstruction emulation. However, it is possible for MCU1 or MCU2 to supply the address to each and every one of the micromemory units 20, 22, 24. The microcode from each micromemory unit controls certain CPU resources.
When a particular macroinstruction requires the use of the operand derivation controller or the execution phase controller to have access to a particular CPU resource not normally under its control, the microcode from the micromemory units will request the address selection unit associated with the micromemory unit controlling the desired CPU resource to cause that address selection unit to select the 10-bit address from the microcontroller requesting the additional CPU resource. If the CPU resource is available, the appropriate address will be strobed into the address selection unit and forwarded to the associated micromemory unit. If however, the address selection unit receiving the request is currently being used by the other microcontroller, the address selection unit receiving the request will return a "hold" signal to the address selection unit associated with the requesting microcontroller. This hold signal, A HOLD or B HOLD, is inputted to either WAIT 1, WAIT 2, or WAIT 4 of the ASU associated with the requesting microcontroller. If a wait signal is received back from a requesting ASU unit, the ASU unit will be inhibited from generating a microcycle clock signal for the computer resources under control of the ASU unit. In other words, that ASU unit goes to sleep until the hold signal is dropped indicating that the CPU resources are available for the requesting microcontroller unit to continue its instruction emulation sequence. As soon as the CPU resource becomes available, the HOLD signal will be negated permitting the requesting microcontroller unit's address to be inputted to the address selection units to bring out the next microcoded instruction needed to control the desired CPU resource.
Still referring to FIG. 2, the clock signals generated by the ASU units, left clock (LC), middle clock (MC) and right clock (RC) are shown. These three clock signals are supplied to the various CPU resources controlled by the microcode stored in the micromemory unit associated with the ASU unit generating the clock. In certain computer macroinstructions, such as a 32-bit add, an increase in one of the phases of the generated microcycle clock signal is needed in order to provide certain CPU resource circuits enough time to perform the extended arithmetic function. Inputs AP1, BP1, AP2 and BP2 are provided to permit microcode control of the length of the two clock phases, Phase 1 (P1) or Phase 2 (P2).
Referring once again to FIG. 1, a catenated connection of the bus interface unit (BIU4-1) is shown for interfacing the input/output data busses of various functional blocks of the invention and for connecting main memory to the CPU. The output bus (system bus) from BIU4-1 contains the address and data bus lines to the main memory. As shown in FIG. 13, a similar arrangement of bus interface units is shown for connecting the address and data bus lines to the main memory interface and to a peripheral interface. Each BIU is identical in circuit design, with control-protocol signals controlling the function of the unit to be bidirectional (send) or unidirectional (receive).
Each microinstruction received and applied to the B0 output bus lines of the BIU 58, 60, 62 and 64 is applied as an address to the op-code transform ROM 36. The op-code transform ROM 36 strips the operand derivation and instruction execution information from each macroinstruction to generate two 8-bit addresses to the operand derivation MCU2 34 and the instruction execution MCU1 32.
To perform the logical operations on the digital data of the computer, general processing means 41 and 43 are provided. General processing means 41 is associated with the MCU1 32 for performing the instruction execution phase of the emulation sequence, while the general processing means 43 is associated with the MCU2 34 for performing the operand derivation phase of the emulation. For the general processing means 41, general processor units (GPU) 46, 48, 50, 52 are connected for 32-bit operation. Each GPU 46, 48, 50, 52 is identical, and comprises an 8-bit slice processor. Associated with GPU 46, 48 is a register select unit 2 (RSU2) 42. Register select unit 1 (RSU1) 44 is associated with GPU 50, 52. The general processing means 43 associated with MCU2 34 comprises only GPU 54, 56 connected for 16-bit operation. Control of the general processing means 43 is directly from macrocode while control of the GPU units of the general processing means 41 are supplied by the RSU units 42, 44.
Still referring to FIG. 1, the computation section of the CPU emulator 23 contains 2 pairs of General Processor Units (GPUs) 46, 48 and 50, 52. The GPU pairs are referred to as an "upper arithmetic logic unit (ALU)" 46, 48 and a "lower ALU" 50, 52 deriving their designations from the respective most and least-significant halves of 32-bit arithmetic computations. Since each GPU is an 8-bit slice, the dual 16-bit ALU arrangement provides the 32-bit parallel arithmetic capability.
Each GPU pair contains sixteen 16-bit registers. The sixteen general register assignments are identical in the two GPU pairs. In other words, at the end of each macroinstruction, the contents of the registers of the upper ALU are the same as the contents of the respectively addressed registers of the lower ALU.
The dual 16-bit ALU is employed for a number of reasons. First, the ability to configure a 32-bit parallel ALU simplifies the implementation of double-precision shift, double-precision arithmetic and floating-point arithmetic macroinstructions. This simplification is directly reflected in the utilization of less microprogram memory to implement these functions. When performing single precision (16-bit) operations, both GPU pairs 46, 48 and 50, 52 peform the same function at the same time.
The 32-bit parallel ALU capability also enhances the execution speed of many of the double precision macroinstructions. Most affected are the double precision macroinstructions that, once fetched, require more than one microinstruction cycle (microcycle) to perform their assigned function, such as multiplication, division, and floating-point. At the completion of double precision operations, the corresponding register in the opposite GPU pair has to be loaded with the new result. There are data paths within the CPU to accomplish this correction in one microcycle. However, because of the added correction cycle, the execution speed for simple double precision macroinstructions, such as add and subtract, is not improved over a conventional single 16-bit ALU.
A third reason for replicating the general registers in both GPU pairs 46, 48 and 50, 52 is to avoid the need to multiplex the GPU status signals to be loaded into the status register (not shown). With register replication, the carry output, overflow and sign bit are always available at the upper GPU pair regardless of whether a single or a double precision operation is in progress. Also, register selection control can address the double precision pair of registers independent of whether an even or odd register address was referenced. If some of the general registers were in one GPU pair and the rest in the other, then multiplexers would be required to obtain the status signals from the GPU containing the register of interest. With register replication, the zero detect spans the full 32-bits of both GPU pairs. Single precision zero detection is obtained by instructing both GPU pairs to test the same single precision quantity. Although examining both GPU pairs, the zero detect will obtain the appropriate result because both GPU pairs possess the identical result.
Referring momentarily to FIG. 6, the macroinstruction format for a Univac AN/UYK-20 16-bit minicomputer is shown. The preferred embodiment of the invention is disclosed and discussed herein with regard to the emulation of the macroinstruction set for this computer, and is provided merely for purposes of illustrating an example of how the invention functions to emulate another computer.
The two internal GPU register sources for register-to-register operations are addressed by the GPU "R" and "T" inputs. Only the "R" input address references the destination register to be loaded. In most UYK-20 macroinstructions, the destination register is specified by the "a" field. Therefore, the emulator 23 associates the "a" field with the "R" input and the "m" field with the "T" input. Separate register select units (RSU 2 and RSU 1) are used to provide the "R" and "T" inputs to each pair of GPUs 46, 48 and 50, 52, respectively. Each RSU contains two 8-bit registers known as the source register, "SR", and the operand register "OR". When a new macroinstruction is processed, the "a" and "m" fields are loaded into both the "SR" and the "OR" registers of both RSUs.
During normal operation, the source of the "R" and "T" outputs are defined by a 4-bit internal RSU register called the register select register. The register select register is loaded via the macroinstruction through the RSU MF inputs. The function controlled by each MF bit is as follows:
______________________________________Source ofRegisterSelectRegisterBit Function When Low When High______________________________________MF3 "R" Output "SR" bits 70-4 "OR" bits 3-0 MultiplexerMF2 "R" Output No Increment Increment IncrementerMF1 "T" Output "S" bits 3-0 "OR" bits 7-4 MultiplexerMF0 "T" Output No Increment Increment Incrementer______________________________________
Therefore, by loading the lower half of the macroinstruction into both the "SR" and "OR", the microprogrammer can select (a,m), (a,a), (m,a), or (m,m) for the GPU register select inputs.
The RSU also has a DEFAULT input to force the register select register outputs to a known state without changing its contents. When activated, the outputs are forced low resulting in the selection of the four most-significant bits of "SR" to the "R" output and the four least-significant bits of "SR" to the "T" output. Neither select output is incremented. Since "SR" is loaded with the macroinstruction "a" and "m" fields in its four most and least significant bit positions, respectively, the default is used to quickly provide "a" at "R" and "m" at "T" independent of the contents of the register select register. This feature allows for the storage of register select controls for double precision and the use of the DEFAULT line for single precision.
RSU1 and RSU2 have the ability, under microcontrol, to displace the register being addressed in the GPU by one count or position. In other words, the register being addressed in the most significant GPU pair 46, 48 would be "a" and the register being addressed in the second GPU pair 50, 52 would be "a+2". This ability to increment the register address loaded from the instruction field of the macroinstruction is built into the RSU units. This structure, while adding several circuits, provides several features that result in a more efficient architecture.
With the ability to offset the address of the selected register and with replication of the data stored in the registers, the ALU is always set up for immediate double precision operations. Also, the structure is not dependent on having even numbered register assignments referenced in double precision instructions. When single precision instructions are executed, the operation is duplicated in both GPU pairs. A very important feature of this approach is that all arithmetic status information to the microcontroller is generated from the same place for both single and double precision operations. No additional circuitry or operations are required to combine sequentially generated arithmetic status. The only overhead associated with this approach is that after a double precision operation, the register in each GPU pair that changed in the other pair has to be updated to the new value. The cross switching in the BIU units 58, 60, 62, 64 allows for the simultaneous update of both GPU pairs 46, 48 and 50, 52 in one microcycle clock time.
As previously mentioned, the dual GPU pairs represent a reconfigurable 16-bit/32-bit ALU. What is referred to as the accumulator extension register is contained within the BIU units. This extension register is used for other purposes to be described later. The most significant outputs of the shifter in the BIU units are connected to the least significant outputs of the lower GPU pair 50, 52 shift out. Now with the proper placement and zero fill (for single precision) of the multiplicand or dividend in the ALU and using the BIU extension register to hold the multiplier and divisor, the same basic multiply and divide microcommands are used for all multiply and divide executions (single precision, double precision, and floating point).
As mentioned earlier, one of the advantages in the design of the invention is the minimization of micromemory resulting in power savings. The structure and control of the ALU is, therefore, designed with the intent that only one series of microcode is required to execute all identical operations in which the only differences are precision or operand derivation. The functional requirements of the RSU and BIU circuits are integral to the ALU and provide for the efficient handling of these features without paying a severe time penalty.
Still referring to FIG. 1, four concatenanted BIU circuits (see discussion in section entitled "The Bus Interface Unit") form the two bidirectional to unidirection interfaces that connect the CPU function to the system bus. Each interface has a temporary register (TA & TB) to allow the interface to complete data transfers on the bus without continuous CPU control. The BIU also contains a shifter which uses TA as the source and TB as the slave. For the multiply and divide implementations, the BIU registers and shifter are used for the ALU extension register. Bus transfers would not be required by the CPU while performing the multiply and divide iteration sequences.
Connection is made from the most significant output of the BIU to the least significant input of the lower ALU. For multiply, the multiplier is loaded into the BIU register. The operation to be performed in the ALU to form the product is determined by looking at the LSB of the BIU shift out. For divide, the least significant portion of the dividend is loaded into the BIU register. As it is shifted into the ALU dividend register, the bits forming the quotient are shifted in behind it.
By extending the multiplicand and division with zeros, the single precision multiply and divide iterations can be done with the ALU configured for 32-bit parallel operation. With this configuration the same basic multiply and divide operation microcode can be used for both single and double precision executions. The extension is also used for floating point alignment and normalization. Also contained in the BIU are multiplexers to move data from the unidirectional input on one side to the unidirectional output on the other. Therefore, the extension of CPU data paths to allow the single cycle double precision register updates, and parallel inputs to the ALUs for single precision are contained in the BIU.
The emulator 23 architecture shown in FIG. 1 incorporates two independent microcontrollers, MCU1 32 and MCU2 34. The approach incorporates several concepts designed for more efficient utilization of micromemory 21 and at the same time maintain macroinstructions speeds for register-to-register operations. The dual-controller approach is based upon the observation that performance of most macroinstructions can be accomplished with two sequences of microcode, both of which can be independently common to other macroinstructions. One sequence is called classification or operand derivation. Macroinstructions are classified by whether they use a second general register, a general register pair, a macroinstruction literal field, a bit-position mask, a memory word, or a memory double word as an operand source. In most cases this could be called the operand derivation. The other sequence is by mathematical function. Among these functions are load, add, subtract, one's complement, negate (two's complement), compare, and logical AND and OR. This sequence is referred to as execution.
A macroinstruction being input to the CPU first passes through an op-code transform ROM 36. The output is two micromemory entry points. The first is to the classification controller and the other is to the execution controller. Both controllers proceed to access micromemory based on their respective entry points.
The micromemory 21 is divided into three sections, LM 20, MM 22 and RM 24. The ASU units 26, 28 and 30 provide the means by which each section can be allocated to either controller in a well defined manner. One section, RM 24, is primarily addressed by the classification controller MCU2 34. This section contains most of the control fields to the circuits that are used mainly for fetch, operand derivation, and interrupt control. Another section LM 20, is primarily addressed by the execution controller, MCU1 32. This section contains most of the common controls necessary to perform function execution of the macroinstruction. The third section is in the middle, MM 22, and control of it is passed from the MCU2 34 when operand derivation is complete to MCU1 32 to finish function processing. The middle section provides the missing control signals indicating where to get the operand.
This architecture is designed to give optimum performance with no duplication of microcode. The same microcode that performs a single precision add is used for all corresponding addressing and data formats. This approach also allows for better packing of microcode. As previously mentioned, in conventional prior-art architectures with single micromemory address, the controls associated with execution would remain a no-op while operand derivation was occurring, and the complement would hold true during execution. With sectioned micromemory much of the respective microcode can occupy the same address space and operand derivation and function execution can proceed in parallel and at the same time.
The program being executed in each Controller, MCU1 32 and MCU2 34, may at times require the entire processor to perform its task. For example, MCU2 34 needs access to general registers located in the arithmetic computation GPUs for use as index registers. In the other direction, MCU1 32 needs access to temporary registers located in the address GPUs. For this reason, each portion of the microprogram memory must be available to both MCUs. The means by which the MCUs request and receives control of the sections of micromemory is incorporated in the address select units (ASU1-3) 26, 28 and 30. Refer to the section entitled "The Address Selection Unit" for a more detailed discussion of these units.
The AN/UYK-20 computer has instructions referencing bit, literal, and byte data. The formating of this data for input to the ALU is accomplished in the RSU circuits. The data referenced by all of these operations is contained in the lower 8 bits of the macroinstruction which is loaded into both RSUs 42, 44 at the beginning of each macroinstruction cycle.
Each RSU contains all of the source data and can output eight bits of the formatted data. Each RSU can be selected to output either the most or least significant 8 bits of the literal expansion, byte expansion, or 1-of-16 bit mask. The two RSUs 42, 44 operating together generate the full 16-bit operand.
Referring now to both FIGS. 1 and 3, each GPU 46, 48, 50, 52, 54, 56 is an 8-bit bit-slice processor consisting of: a 16-word by 8-bit dual access register file 70, two port buffers (P1B 78, P2B 80), extensive data type selection multiplexing, arithmetic logic circuit (ALC) 86 and powerful shifting capability. It has separate 8 bit data input and data output paths. Data can be input directly to the register file 70, P1B 78 or P2B 80. Data can be output from P1B 78, P2B 80 or from the ALC 86. The data paths are designed to facilitate the implementation of complex algorithms. For example, two bit multiply data paths consisting of a right shift of one bit into the ALC 86, and right shift of one or two bits into the register file 70 are implemented in the unit.
Table 1 contains a description of the I/O signals of each GPU unit. The I/O signals can be classified by function as:
Data (DI, DO, CI, CO, AZO, MXH, MXL)
Addresses (R, T)
Control (A, C, D, M, S, DOE, LC)
Data Out (DO) is an 8-bit parallel tri-state output controlled by the data out enable (active low) input. CO is the carry-out. All zero is an open collector output. MXH and MXL(0) are bidirectional input/output pins. MXL(1) is a tri-state output. The function of the GPU is the result of the coordinated exercise of 10 distinct subfunctions which are described as follows:
TABLE 1______________________________________GPU SIGNAL LISTMnemonic Description______________________________________DO(3) Data Output bit 3DO(4) Data Output bit 4DO(5) Data Output bit 5DO(6) Data Output bit 6DO(7) Data Output bit 7CO Carry OutDOE Data Out Enable (active low)AZO All Zero Detect OutMXH(0) Multiplexer Shift High bit 0MXH(1) Multiplexer Shift High bit 1C0 Boundary, Connect Control bit 0C1 Boundary, Connect Control bit 1C2 Boundary, Connect Control bit 2DI(7) Data Input bit 7DI(6) Data Input bit 6DI(5) Data Input bit 5DI(4) Data Input bit 4DI(3) Data Input bit 3DI(2) Data Input bit 2DI(1) Data Input bit 1DI(0) Data Input bit 0S1 Source Select Control bit 1S0 Source Select Control bit 0R2 Port 1 Register Select bit 2R3 Port 1 Register Select bit 3T2 Port 1 Register Select bit 2T3 Port 1 Register Select bit 3R1 Port 1 Register Select bit 1T1 Port 1 Register Select bit 1T0 Port 1 Register Select bit 0R0 Port 1 Register Select bit 0M1 Destination Select Control bit 1M0 Destination Select Control bit 0M2 Destination Select Control bit 2MXL(0) Multiplexer Shift Low bit 0MXL(1) Multiplexer Shift Low bit 1LC Latch ClockD2 Data Type Select Control bit 2A0 Adder Logic Circuit Control bit 0A1 Adder Logic Circuit Control bit 1C1 Carry InD1 Data Type Select Control bit 1D0 Data Type Select Control bit 0DO(0) Data Output bit 0DO(1) Data Output bit 1DO(2) Data Output bit 2______________________________________
Referring now to FIG. 3, the register file 70 contains 16 words of 8 bits each. Any two words addressed by the addresses R and T can be read simultaneously at port 1 and port 2, respectively. Identical data appears at both ports if R and T are equal. Data can be written into the register file 70 at address R if the load clock (LC) is high. The M-bits determine the source of the data to be written into the register file 70. If M=LLL (L=logic low H=logic high) and S=LH Data In (DI) is written into the register file 70. If M=LLH, LHL or LHH the output of the shifter 68 is written into the register file 70. If M=HLL the load clock is internally suppressed (forced low) and thus no writing into the register file takes place. If M=HLH, HHL or HHH the output of the ALC 86 is written into the register file 70. If the S-bits are HH, the least significant bit of the R-address is internally forced to 1. Table 2 illustrates the source for data loaded into the register file 70.
TABLE 2______________________________________REGISTER FILE INPUT SOURCE SELECTIONM SOURCE NOTES______________________________________LLL DATA IN (DI) S=LH RequiredLLH SHIFTERLHL SHIFTERLHH SHIFTERHLL LC DISABLED-NO WRITEHLH ALCHHL ALCHHH ALC______________________________________
The port 1 buffer (P1B) 78 can operate in 3 modes: as a slave to the register addressed by R (i.e. to port 1), as a falling edge triggered master/slave register with port 1 as input, or as a latch to data in (DI). The mode is determined by the M-bits and by the S-bits. If S=LL, HL or HH, P1B 78 operates as a slave to the register addressed by R. Together Reg(R) and P1B form a falling edge triggered master/slave register. Thus the value of P1B 78 will remain stable while LC=H and REG(R) is being written. When LC=L, P1B will follow the value appearing on port 1. P1B will change value while LC=L, if R is changed.
If S=LH and M=LLL, P1B operates as a master/slave register. Its input is port 1 of the register file 70. In this mode, P1B 70 changes value only with the falling edge of LC. Unlike in the previously described mode, if R is changed while LC=L, P1B will not follow port 1. Note that in this mode DI is being written into the register dile 70. If S=LH and M=LLL, P1B functions as a latch for DI. In this mode P1B follows DI while LC=L and does not change while LC=H, retaining the value of DI at the time of the rising edge of LC.
Mode switching should take place when the load clock is low. If M changes to LLL while S=LH, P1B will retain the value of DI at the time M changed, until the falling edge of LC. If DI and M change simultaneously the new value of DI will be retained. Table 3 illustrates the operating modes for P1B 78.
TABLE 3______________________________________P1B OPERATING MODESS=LH M=LLL MODE______________________________________NO DON'T SLAVE TO REG(T) CAREYES NO DISABLED, VALUE NOT CHANGINGYES YES MASTER/SLAVE REGISTER FOLLOWS REG(R)______________________________________
The Port 2 Buffer P2B 80 can operate in 3 modes: as a slave to the register addressed by T (i.e. to port 2), as a transparent latch for DI, or disabled. If S=LL or LH and A=LL, HL or HH as well as when S=HL or HH and A=LH, P2B 80 functions as a slave to the register addressed by T. Together, Reg(T) and P2B 80 form a falling edge triggered master/slave register. Thus the value of P2B 80 remains stable when LC=H, even if Reg(T) is being written into (if R=T). When LC=L, P2B will follow the value appearing on Port 2. P2B will change value while LC=L, if T is changed.
If S=HL or HH and A=LL, HL or HH, P2B operates as a transparent latch for DI. In this mode P2B 80 always follows DI. If S=LL or LH and A=LH, all inputs to P2B 80 are disabled and P2B 80 retains the value it held when it was disabled, until the mode is changed. In this manner values appearing on port 2 or on DI can be stored indefinitely in P2B 80. Table 4 illustrates the operating modes for port P2B 80.
TABLE 4______________________________________P2B OPERATING MODESS A MODE______________________________________LL, LH LL, HL, HH SLAVE TO REG(T)LL, LH LH DISABLED, VALUE NOT CHANGINGHL, HH LL, HL, HH FOLLOWS DI UNCONDITIONALLYHL, HH LH SLAVE TO REG(T)______________________________________
Still referring to FIG. 3, the left data type selector 82 supplies the left operand to the ALC 86. The left operand can be: P1B 78, the 1's complement of P1B 78 (P1B), zero, or P1B 78 shifted right 1 bit. The left data type selector 82 is mainly controlled by the D-bits. The A-bits and C-bits enter into consideration in some cases. The left operand of the ALC 86 is P1B 78 when D=LLH, HLH or HHH. When D=LHH the left operand of the ALC 86 is P1B. Zero is used as the left operand of the ALC when D=LLL or HLL. If D=LHL or HHL the left operand of the ALC 86 is P1B if A=LL, HL or HH and is P1B 78 shifted right 1 bit if A=LH.
In the case where the left operand of the ALC 86 is P1B 78 shifted right 1 bit, the most significant bit is controlled by the C-bits. For C=LLL or LLH, the most significant bit of the left ALC 86 operand equals the input value applied to the pin MXH(1). For C=HLL a sign extended shift is used, i.e., the most significant bit equals bit 7 of P1B. If C=HLH, the most significant bit of the left of the ALB 86 operand is equal to the value of the flip-flop TS. If C=HHL or HHH the most significant bit of the left ALC 86 operand is equal to bit 7 of P2B 80. The values C=LHL and LHH should be avoided when the left ALC 86 operand is P1B shifted right 1 bit, because the value of the most significant bit of the left ALC 86 operand is undefined for this control setting.
The right data type selector 84 supplies the right operand to the ALC 86. The right operand can be P2B 80, the 1's complement of P2B (P2B), or sero. The right ALC 86 operand is determined by the D-bits as follows. The right ALC operand is P2B if D=HLL, HLH or HHL. It is P2B if D=LLL, LLH or LHL, and it is zero if D=LHH or HHH. Table 5 illustrates the ALC 86 operands from the left and right data type selectors 82, 84 for the various conditions of the D and A bits. Table 6 illustrates the most significant bit when 1/2 of the P1B part 78 is input to the ALC 86.
TABLE 5______________________________________ALC OPERANDSD A=LH LEFT RIGHT______________________________________LLL Don't Care 0 ##STR1##LLH Don't Care P1B ##STR2##LHL No ##STR3## ##STR4##LHL Yes 1/2 * P1B ##STR5##LHH Don't Care ##STR6## 0HLL Don't Care 0 P2BHLH Don't Care P1B P2BHHL No ##STR7## P2BHHL Yes 1/2 * P1B P2BHHH Don't Care P1B 0______________________________________
TABLE 6______________________________________MSB FOR 1/1 * P1B AS LEFT ALC OPERAND C MSB______________________________________ LLL MXH(1) LLH MXH(1) LHL Undefined LHH Undefined HLL P1B(7) HLH TS HHL P2B(7) HHH P2B(7)______________________________________
The ALC 86 can perform addition and the logical operations AND and OR on the operands supplied by the left and right data type selectors 82, 84, respectively. A carry-in can be externally supplied by means of the input pin CI. In addition to the 8-bit result, the ALC 86 produces a carry-out on the output pin CO, an overflow indication which can be routed to MXH(1) if C=LHL or C=LHH, and an all-zero signal which causes the AZO pin to be in a high impedance state when the ALC result is zero, and to be grounded if it is not zero.
The function of the ALC 86 is controlled by the A-bits. The ALC 86 adds when A=LL or A=LH, it performs AND if A=HL, and if performs OR when A=HH. The carry-out is computed using carry look-ahead. For both logical operations the carry-out equals the carry-in. Overflow is computed as the exclusive-OR of the carry into and out of the most significant bit. Table 7 illustrates the ALC 86 functions as a function of the A(1-0) bits.
TABLE 7______________________________________ALC 86 FUNCTIONS A FUNCTION______________________________________ LL ADD LH ADD HL AND HH OR______________________________________
The Shifter 68 is capable of shifting the output of ALC 86 one bit to the right, two bits to the right or one bit to the left. The M-bits select one of the these 3 shifts. If M=LLH the output of the shifter 68 is the output of ALC 86 shifted left 1 bit. If M=LHL the output of the shifter 68 is the output of ALC 86 shifted right 1 bit. If M=LHH the output of the shifter 68 is the output of the ALC 86 shifted right 2 bits. For all other values of the M-bits the function of the shifter is of no consequence, as no output or storage element state depends on it. Table 8 illustrates the function of the shifter 68 as a function of the M(2-0) bits.
TABLE 8______________________________________SHIFTER 68 FUNCTIONSM FUNCTION______________________________________LLH SHIFT ALC OUTPUT LEFT 1 BITLHL SHIFT ALC OUTPUT RIGHT 1 BITLHH SHIFT ALC OUTPUT RIGHT 2 BITS.______________________________________
The value of the shift carries (i.e., the values entered into the bit positions left vacant as a result of a shift operation) is controlled mainly by the C-bits. For left shifts of 1 bit, the least significant bit is set to 0 if C=LHH, to 1 if C=LHL, and set equal to the value of MXL(0) (which is in input mode) for any other value of the C-bits. Table 9 illustrates the shifter carry for left shifts of 1 bit in the shifter 68 as a function of the C(2-0) bits.
TABLE 9______________________________________SHIFT-CARRY FOR LEFT SHIFTS OF 1 BITC VALUE OUTPUT BY SHIFTER BIT 0______________________________________LLL MXL(0)LLH MXL(0)LHL 1LHH 0HLL MXL(0)HLH MXL(0)HHL MXL(0)HHH MXL(0)______________________________________
For right shifts of 1 bit, the most significant bit is set to 0 if C=LHH, and to 1 if C=LHL. If C=LLL or C=HLL the most significant bit is set to the value of bit 7 of the ALC 86 output, unless there is an overflow in the ALC; in that case the most significant bit is set to the complement of bit 7 of the ALC 86 output. If C=LLH, HLH, HHL or HHH the most significant bit is set to the value of MXH(0) (which is in input mode). Table 10 illustrates the shift-carry for right shifts of 1 bit in the shifter 68 as a function of the C(2-0) bits.
TABLE 10______________________________________SHIFT-CARRYFOR RIGHT SHIFTS OF 1 BITC VALUE OUTPUT BY SHIFTER BIT 7______________________________________LLL ALC(7) ⊕ OVFLLH MXH(0)LHL 1LHH 0HLL ALC(7) ⊕ OVFHLH MXH(0)HHL MXH(0)HHH MXH(0)______________________________________
For right shifts of 2 bits, the two most significant bits are set to 0 if C=LHH, and to 1 if C=LHL. If C=LLL or C=HLL, the two most significant bits are set to the value of bit 7 of the ALC 86 output, unless there is an overflow in the ALC; in that case the two most significant bits are set to the complement of bit 7 of the ALC output. If C=LLH, bit 7 is set equal to the input value on MXH(1) and bit 6 is set equal to the input value of MXH(0). If C=HLH, bit 7 is set equal to the value of the flip-flop TS and bit 6 is set equal to the input on MXH(0). If C=HHL or C=HHH, bit 7 is set equal to bit 7 of P2B 80 and bit 6 is set equal to the input on MXH(0). Table 11 illustrates the shift-carry for right shifts of 2 bits in the shifter 68 as a function of the C(2-0) bits.
TABLE 11______________________________________SHIFT-CARRYFOR RIGHT SHIFTS OF 2 BIT Value Output by ShifterC Bit 7 Value Output by Shifter Bit 6______________________________________LLL ALC(7) ⊕ OVF ALC(7) ⊕ OVFLLH MXH(1) MXH(0)LHL 1 1LHH 0 0HLL ALC(7) ⊕ OVF ALC(7) ⊕ OVFHLH TS MXH(0)HHL P2B(7) MXH(0)HHH P2B(7) MXH(0)______________________________________
The output values appearing on the pins MXH(1), MXH(0), MXL(1) and MXL(0) are controlled mainly by the C-bits and the M-bits. MXH(1) is in input mode if C=LLL or C=LLH. If C=LHL or C=LHH the overflow signal from the ALC 86 is output on MXH(1). If C=HLL, the value of bit 6 of the ALC 86 output is output on MXH(1). If C=HLH, the value output on MXH(1) is the value of the flip-flop TS 67. If C=HHL or C=HHH, bit 7 of P2B 80 is output on MXH(1). Table 12 illustrates the MXH(1) function as a function of the C(2-0) bits.
TABLE 12______________________________________MXH(1) FUNCTIONC FUNCTION______________________________________LLL INPUTLLH INPUTLHL OUTPUT OVFLHH OUTPUT OVFHLL OUTPUT ALC(6)HLH OUTPUT TSHHL OUTPUT P2B(7)HHH OUTPUT P2B(7)______________________________________
MXH(0) is in input mode if C=LLL or if M=LHL or M=LHH. At all other times the value of bit 7 of the ALC 86 output is output on MXH(0).
MXL(1) is in output mode if M=LHH or if A=LH and D=LHL or D=HHL. If A=LH and either D=LHL or D=HHL, the value of bit 0 of P1B 78 is output on MCL(1). If A=LH or D=xLH, but M=LHH, the value of bit 1 of the ALC 86 output is output on MXL(1). If MXL(1) is not in output mode, it is placed into a high impedance state. Table 13 illustrates the MXL(1) function as a function of the A, D and M bits.
TABLE 13______________________________________MXL(1) FUNCTION(A=LH)Λ(D=xHL) M=LHH FUNCTION______________________________________NO NO HIGH Z STATENO YES OUTPUT ALC(1)YES NO OUTPUT P1B(0)YES YES OUTPUT P1B(0)______________________________________
MXL(0) is in input mode unless M=LHL or M=LHH. If M=LHL or LHH, the value of bit 0 of the ALC 86 output is output on MXL(0).
The GPU can output 8 bits in parallel on the data output (DO) bus. There is one tri-state driver for each bit in DO. All 8 drivers can be disabled by a high input on the data out enable pin (DOE).
One of three values can be output on DO: the output of the ALC 86, the output of P1B 78, or the output of P2B 80. The M-bits control which of these is output on DO. If M=HHH, P1B 78 is output on DO. If M=HHL, P2B 80 is output on DO. For any other value of M, the output of the ALC 86 is output on DO.
Bit 7 of the ALC 86 output can be latched into the D flip-flop TS 67. TS 67 is enabled for input whenever C=HHH.
The function performed by the GPU is determined by the set of 15 control inputs. The effects of the various settings of these control signals are discussed in the following sections.
The two S-bits affect the function of the register file 70, of P1B 78, and of P2B 80. The register file 70 is affected by the S-bits in two ways. First, when S=HH, the least significant bit of the R address is forced to 1 and thus only odd-address registers can be accessed. Second, to write incoming data from data in (DI) into the register file 70, the S-bits must be set to LH (in addition to setting M=LLL).
The operating mode of the port 1 buffer 78 is determined by the value of the S-bits and M-bits. If S=LL, HL or HH, P1B 78 operates as a slave to Reg(R). If S=LH and M=LLL, P1B 78 operates as a latch for data in (DI). If S=LH and M=LLL, P1B 78 operates as a falling edge triggered master/slave register and followed Reg(R).
The operating mode of P2B 80 depends on the most significant of the two S-bits (S1) and on the A-bits. If S1 =L and A=LH of if S1 =H and A=LH, P2B 80 operates as a slave to Reg(T). If S1 =L and A=LH, P2B 80 is disabled in the sense that its value does not change until either S1 or the A-bits change. If S1 =H and A=LH, P2B 80 operates as a transparent latch for data in (DI) in the sense that the output of P2B 80 follows DI unconditionally. Table 14 illustrates the effects of the S-bit settings.
TABLE 14__________________________________________________________________________EFFECTS OF S-BIT SETTINGSS REGISTER FILE P1B P2B__________________________________________________________________________LL M≠LLL No Effect SLAVE TO REG(R) A≠LH SLAVE TO REG(T) M=LLL Write Inhibit A=LH DISABLEDLH M≠LLL No Effect M≠LLL Latch for DI A≠LH SLAVE TO REG(T) M=LLL Write Enable M=LLL Master/Slave Mode A=LH DISABLEDHL M≠LLL No Effect SLAVE TO REG(R) A≠LH FOLLOWS DI M=LLL Write Inhibit A=LH SLAVE TO REG(T)HH M≠LLL R(0) ← 1 SLAVE TO REG(R+1) A≠LH FOLLOWS DI M=LLL Write Inhibit A=LH SLAVE TO REG(T)__________________________________________________________________________
The Three D-bits affect the function of the data type selectors and MXL(1). The left data type selector 82 outputs the value of P1B 78 if D=LLH, HLH or HHH. It outputs zero if D=LLL or LLH. The output of the left data type selector 82 is P1B 78 either if D=LHH or alternately if D=xHL (x means "don't care") and A=LH. If D=xHL and A=LH the output of the left data type selector 82 is P1B 78 shifted right 1 bit. In the latter case the value of the most significant bit of the left data type selector 82 output is determined by the C-bits.
The right data type selector 84 outputs P2B 80 if D=HLL, HLH or HHL. If D=LLL, LLH or LHL the output is P2B 80. Zero is output by the right data type selector 84 if D=LHH or HHH. The output pin MXL(1) outputs bit 0 if P1B 78 if D=xHL and A=LH. If the D-bits have any value other than LHL or HHL (or if A=LH), the M-bits determine the function of MXL(1). If M-LHH, MXL(1) outputs bit 1 of the ALC 86 output. Otherwise MXL(1) is in a high impedance state. Table 15 illustrates the effects of the D-bit settings.
TABLE 15______________________________________EFFECTS OF D-BIT SETTINGS Right Data Left Data TypeD Type Select Select MXL(1)______________________________________LLL 0 ##STR8## M≠LHH M=LHH HIGH Z ALC(1)LLH P1B ##STR9## M≠LHH M=LHH HIGH Z ALC(1)LHL A≠LH A=LH ##STR10## ##STR11## M≠LHH A≠LH A=LH HIGH Z P1B(0) M=LHH ALC(1)LHH ##STR12## 0 M≠LHH M=LHH HIGH Z ALC(1)HLL 0 P2B M≠LHH HIGH Z M=LHH ALC(1)HLH P1B P2B M≠LHH HIGH Z M=LHH ALC(1)HHL A≠LH A=LH ##STR13## P2B M≠LHH A≠LH A=LH HIGH Z P1B(0) M=LHH ALC(1)HHH P1B 0 M≠LHH HIGH Z M=LHH ALC(1)______________________________________
The two A-bits affect the function of the ALC 86, P2B 80, the left data type selector 82 and of MXL(1). The function performed by the ALC 86 is determined by the A-bits. If A=LL or LH the ALC 86 adds. If A=HL the logical operation AND is performed in the ALC, and if A=HH the logical operation OR is performed in the ALC. When the ALC adds the carry-out CO is computed by carry look-ahead. For logical operations the carry-out CO equals the carry-in CI. Overflow (which may be output on MXH(1) and which plays a role in both left shift-carries) is always zero when logical operations are performed.
The operating mode of P2B 80 is determined by the A-bits and by the most significant of the two S-bits (S1). When A=LH, P2B 80 operates as a slave to Reg(T) if S1 =L, and the output of P2B follows DI unconditionally if S1 =H. When A=LH, P2B is disabled as long as S1 =L, and it acts as a slave to Reg(T) if S1 =H. The function the left data type selector 82 depends on the value of A-bits only when D=LHL or HHL. In these two cases, the output of the left data type selector is P1B if A=H, and to the output is P1B 78 shifted right on ebit if A=LH.
The output pin MXL(1) outputs bit 0 of P1B 78 if A=LH and D=xHL. If the A-bits have any other value (or if D=xHL), the M-bits determine the function of MXL(1). if M=LHH, MXL(1) outputs bit 1 of the ALC 86 output. Otherwise MXL(1) is in a high impedance state.
The three M-bits affect the function of the data output (DO), of the shifter 68, of the MXH and MXL pins, of the register file 70 and of P1B 78. The load clock (LC) is internally disabled if M-HLL. The data output is a tri-state 8-bit set of output signals. If DOE=H all 8 pins are placed in a high impedance state. If DOE=L the M-bits determine the value to be output on DO. If M=HHH the value of P1B 78 is output. If M=HHL the value of P2B 80 is output. For all other values of the M-bits the output of ALC 86 is output through DO. Table 16 illustrates the effect of the A bits, while Table 17 illustrates the effects of the M-bits.
TABLE 16__________________________________________________________________________EFFECTS OF THE A-BITSA ALC P2B Left Data Type Select MXL(1)__________________________________________________________________________LL ADD S=Lx Slave to Reg(T) D≠xHL No Effect M≠LHH High Z S=Hx Follows Dl Uncond. D=xHL ##STR14## M=LHH ALC(1)LH ADD S=Lx Disabled D≠xHL No Effect D≠xHL M≠LHH High Z M=LHH ALC(1) S=Hx Slave to Reg(T) D=xHL 1/2 * P1B D=xHL P1B(0)HL AND, CO=Cl, OVF=0 S=Lx Slave to Reg(T) D≠xHL No Effect M≠LHH High Z S=Hx Follows Dl Uncond. D=xHL ##STR15## M=LHH ALC(1)HH OR, CO=Cl, OVF =0 S=Lx Slave to Reg(T) D≠xHL No Effect M≠LHH High Z S=Hx Follows Dl Uncond. D=xHL ##STR16## M=LHH ALC(1)__________________________________________________________________________
TABLE 17__________________________________________________________________________EFFECTS OF THE M-BITS Source of Write-DataM DO* SHIFTER MXL(1) MXL(0) MXH(0) LC for Register P1B__________________________________________________________________________ MODELLL ALC D≠xHL High Z INPUT C≠LLL ALC(7) Enabled S≠LH Write Inhibit S≠LH Slave or to A≠LH Reg(R) D=xHL P1B(0) C=LLL INPUT S=LH Dl S=LH Master/ and Slave A=LH to Reg(R)LLH ALC Shift left D≠xHL High Z INPUT C≠LLL ALC(7) Enabled SHIFTER S≠LH Slave 1 bit or to A≠LH Reg(R) D=xHL P1B(0) C=LLL INPUT S=LH Latch and for Dl A=LHLHL ALC Shift left D≠HL High Z ALC(0) INPUT Enabled SHIFTER S≠LH Slave 1 bit or to A≠LH Reg(R) D=xHL P1B(0) S=LH Latch and for Dl A=LHLHH ALC Shift right D≠xHL ALC(1) ALC(0) INPUT Enabled SHIFTER S≠LH Slave 2 bits or to A≠LH Reg(R) D=xHL P1B(0) S=LH Latch and for Dl A=HLHLL ALC D≠xHL High Z INPUT C≠ LLL ALC(7) Disabled WRITE INHIBIT S≠LH Slave or to A≠HL Reg(R) D=xHL P1B(0) C=LLL INPUT S=LH Latch and for Dl A=LHHLH ALC D≠xHL High 2 INPUT C≠LLL ALC(7) Enabled ALC S≠LH Slave or to A≠LH Reg(R) D=xHL P1B(0) C=LLL INPUT S=LH Latch and for Dl A=LHHHL P2B D≠xHL High Z INPUT C≠LLL ALC(7) Enabled ALC S≠LH Slave or to A≠LH Reg(R) D=xHL P1B(0) C=LLL INPUT S=LH Latch and for Dl A=LHHHH P1B D≠xHL High Z INPUT C≠LLL ALC(7) Enabled ALC S≠LH Slave or to A≠LH Reg(R) D=xHL P1B(0) C=LLL INPUT S=LH Latch and for Dl A=LH__________________________________________________________________________ *When DOE=H the D0 pins are in a high impedance state.
The function of the shifter 68 is controlled by the M-bits. When M=LLH, the shifter performs a 1 bit left shift. When M=LHL, the shifter performs a 1 bit right shift. When M=LHH, the shifter performs a 2 bit right shift. For all other values of the M-bits the function of the shifter is of no consequence (don't care) since it does not affect the state of any storage element or of any output signal.
The output pin MXL(1) outputs the value of bit 1 of the ALC 86 output if M=LHH and either D=xHL or A=HL. If D=xHL and A=LH then MXL(1) outputs bit 0 of P1B 78. If M=LHH and either D=HL or A=LH then MXL(1) is placed in a high impedance state. The bidirectional pin MXL(0) outputs the value of bit 0 of the ALC 86 output if M=LHL or M=LHH. If M=LHx, MXL(0) is in input mode. The bidirectional pin MXH(0) outputs the value of bit 7 of the ALC 86 output if C=LLL and M=LHx. If C=LLL or M=LHL, MXH(0) is in input mode.
The register file 70 is affected by the M-bits in two ways. First, the M bits determine the source of the data to be written into the register file 70. Second, if M=HLL the load clock (LC) is disabled and thus nothing can be written into the register file. The function of P1B 78 is controlled bythe S-bits and by the M-bits. If S=LH and M=LLL, P1B operates as a master/slave register and retains the value of Reg(R) from the time of the last falling clock edge. If S=LH but M=LLL, P1B operates as a latch for data input (DI). If S=LH, P1B operates as a slave to Reg(R), regardless of the value of the M-bits.
The three C-bits affect the shift-carries into the shifter 68 and into the left data type selector 82 when P1B 78 when P1B 78 is shifted right, as well as MXH(1), MCH(0) and the flip-flop TS 67. The shift-carries are values entered into the bit positions left vacant as a result of a shift operation. For left shifts of 1 bit, the least significant bit is left vacant. It is set to 0 if C=LHH, to 1 if C=LHL, or it is set equal to the value applied at the pin MXL(0) (which is in input mode during left shifts) if C=LHx.
For right shifts of 1 bit, the most significant bit is left vacant. It is set to 0 if C=LHH, or to 1 if C=LHL. If C=LLL or C=HLL the most significant bit is set to the value of bit 7 of the ALC 86 output, unless there is an overflow in the ALC. In that case, the most significant bit is set to the complement of bit 7 of the ALC output. If C=LLH, HLH, HHL or HHH the most significant bit is set to the value of MXH(0) (which is in input mode during right shifts).
For right shifts of 2 bits, the two most significant bits are left vacant. Both are set to 0 if C=LHH and both are set to 1 if C=LHL. If C=LLL or C=HLL both are set equal to bit 7 of the ALC output, unless there is an overflow in the ALC; in that case both bits are set to the complement of bit 7 of the ALC output. If C=LLH, bit 7 of the output of the shifter 68 is set equal to the input value on MXH(1) and bit 6 of the shifter 68 output is set equal to the value of the flip-flop TS 67 and bit 6 of the shifter 68 output is set equal to the input on MXH(0). If C=HHL or C=HHH bit 7 of the shifter 68 output is set equal to the value of bit 7 of P2B and bit 6 of the shifter 68 output is set equal to the input on MCH(0).
In the case when D=LHL or HLL and A=LH, the output of the left data type selector 82 is P1B shifted right 1 bit. If C=LLL or LLH the most significant bit of the left data type selector's 82 output is set equal to value input on MXH(1). For C=HHL the most significant bit of the left data type selector's output is set equal to bit 7 of P1B. For C=HLH the value of the flip-flop TS 67 is used as the most significant bit of the left data type selector's output. If C=HHL or C=HHH the most significant bit of the left data type selector's output is set equal to bit 7 of P2B. The values C=LHL and C=LHH should be avoided when P1B shifted right 1 bit is selected in the left data type selector, because the value of the most significant bit of the left data type selector's output is undefined and unpredictable for this setting of the control bits.
The bidirectional pin MXH(1) is controlled entirely by the C-bits. If C=LLL or C=LLH, MXH(1) is in input mode. If C=LHL or C=LHH, the overflow signal from the ALC is output on MXH(1). If C=HLL bit 6 of the ALC output is output on MXH(1). If C=HLH the value of the flip-flop TS is output on MXH(1). If C=HHL or HHH bit 7 of P2B is output on MXH(1). Table 18 illustrates the effects of the C-bits.
TABLE 18__________________________________________________________________________EFFECTS OF THE C-BITSShift Carries Left DataLeft 1 Bit Right 1 Bit Right 2 Bits Type Select*C Bit 0 Bit 7 Bit 7 Bit 6 Bit 7 MXH(1) MXH(0) TS__________________________________________________________________________LLL MXL(0) ALC(7).OVF ALC(7).OVF ALC(7).OVF MXH(1) INPUT INPUT DISABLEDLLH MXL(0) MXH(0) MXH(1) MXH(0) MXH(1) INPUT M≠LHx ALC(7) DISABLED M=LHx InputLHL 1 1 1 1 UNDEF. OVF M≠LHx ALC(7) DISABLED M=LHx InputLHH 0 0 0 0 UNDEF. OVF M≠LHx ALC(7) DISABLED M=LHx InputHLL MXL(0) ALC(7).OVF ALC(7).OVF ALC(7).OVF P1B(7) ALC(6) M≠LHx ALC(7) DISABLED M=LHx InputHLH MXL(0) MXH(0) TS MXH(0) TS TS M≠LHx ALC(7) DISABLED M=LHx InputHHL MXL(0) MXH(0) P2B(7) MXH(0) P2B(7) P2B(7) M≠LHx ALC(7) DISABLED M=LHx InputHHH MXL(0) MXH(0) P2B(7) MXH(0) P2B(7) P2B(7) M≠LHx ALC(7) ENABLED M=LHx Input__________________________________________________________________________ *Applicable only when D=xHL and A=LH.
The bidirectional pin MXH(0) is in input mode if C=LLL of if M=LHx. Otherwise bit 7 of the ALC output is output on MXH(0). The flip-flop TS 67 is controlled by the C-bits. If C=HHH the value of TS 67 follows bit 7 of output of ALC. For any other value of the C-bits, TS 67 retains its value.
The control bit DOE serves to enable/disable the tri-state outputs for the 8 data output (DO) bits. If DOE=H the tri-state drivers are placed in a high impedance state. If DOE=L data is being output by the tri-state drivers for DO.
The load clock controls the operation of the register file 70, of P1B 78 and of P2B 80. The load clock controls writing of the register file 70. As long as LC=H, the register selected by the R-address (Reg(R)) follows the input to the register file 70. When LC=L all registers remain unchanged. Writing of the register file can be prevented when LC=H by setting M=HLL. Also if M=LLL and S=LH is specified, no writing will take place.
The register file 70 alone is a level mode storage array. The register file 70, however, functions in conjunction with the buffers P1B and P2B. In conjunction with P1B and P2B, the registers selected by the addresses R and T form master/slave registers. When P1B operates as a slave to Reg(R), Reg(R) is write-enabled only when LC=H while P1B is write-enabled only when LC=L, the net effect being that Reg(R) and P1B form a falling edge triggered register.
P2B followed Reg(T) when LC=L. When LC=H, however, P2B is disconnected from port 2 and thus retains its value while LC=H. The net effect is that Reg(T) and P2B form a trailing edge triggered register. When R=T this of no consequence since in that situation Reg(T) cannot be written into. However, when R=T, Reg(T) may change while LC=H, but P2B only follows when the falling edge of LC occurs.
When P1B and P2B operate in the mode of slaves to Reg(R) and Reg(T) respectively, their value could change while LC=L due to a change in the address R or T respectively. When P1B is used in the master/slave mode, its value can only change immediately following the falling edge of the Load Clock. Even if R changes, P1B will then not change until the next falling edge of the load clock. When P1B operates as a latch for DI, P1B follows DI as long as LC=L and retains its value while LC=H.
Referring now to FIG. 4, a block diagram of a RSU unit is shown. The primary functions provided by the RSU circuit are: (1) provide GPU register file select inputs from instruction file or micromemory; (2) generate BIT masks; (3) expand literal and byte data formats; (4) iterate counts outside microcontroller (shift); (5) register select increment and compare for load and store multiple operations; and (6) emit constants (data or address) from micromemory. The RSU has three sets of 8-bit data input ports. The first source is the bus input (BI) port. The second input source is the bus input-output (BIO) port, which not only can receive data from the associated 8-bit, but also can transmit data as well by means of tri-state drivers. The third input data source is the microfield (MF) which can be used to enter constants from the micromemory when not needed for internal control of the RSU.
The data source selection and internal operations are determined by five microcontrol bits. The bus input load (BIL) bit commands that one of the bus input sources be loaded. The bus select designator (BSD) determines which bus is selected and the bus load designator (BLD) determines where the data are to be stored. The micro-field (MF) is selected as data when the micro-field input enable (MFIE) bit is true and the micro-command (MC) bit is false.
The RSU has two internal 8-bit registers that can be loaded in 4-bit groups. The source register (SR) 100 normally will contain the R and T register designators for the GPU(s) associated with an RSU. The upper 4-bits in the SR (SR7-4) normally supplies the R field for the GPU and the lower 4-bits of the SR (SR3-0) normally supplies the T field.
The second internal RSU register is the operations register (OR) 102. The (OR) may be used as a temporary register for transferring constants from micromemory to a GPU, as a shift count register, as an alternate source for the register select outputs, and as a split counter where the upper 4-bits are incremented until they are equal to the lower 4-bits. By proper selection of micro-field bits, either half of the SR 100 may be loaded with the incremented or decremented OR 102 values at the same time that the new values are loaded into the OR.
The RSU also contains a temporary register which holds the updated OR value for one-half clock period. This register is used only to isolate the previous and next values of the OR and cannot be used for additional internal storage.
The R and T outputs of the RSU are each 4-bit outputs intended to control the register selection inputs to a bank of CPUs. Each output set (4 bits) has a 2:1 multiplexer and a half-adder. The (R3-0) outputs can present SR7-4, (SR7-4)+1, OR3-0, or (OR3-0)+1 to the GPU set. The T3-0 outputs can present SR3-0, (SR3-0)+1, OR7-4, or (OR7-4)+1 to the GPU set. The 4-bit code which selects 2 of the above 8possibilities is stored in the register select register by a microcode command (LDRS). The register select control word may be overridden but not lost by a logic 1 on the default input. In the default mode the R outputs reflect SR7-4 and the T outputs reflect SR3-0.
The BIO outputs are made active by a bus output microcommand. The Bus Output command has 6 possible modes which are:
(a) Decode OR3-0 and output a one-of-eight decode for values of OR3-0 between 0 and 7 inclusive.
(b) Decode OR3-0 and output a one-of-eight decode for values of OR3-0 between 8 and 15 inclusive.
(c) Output OR7-0 on BIO7-0.
(d) Output OR3-0 on BIO3-0 and output logic zero on BIO7-4.
(e) Output logic zeros on BIO7-0.
(f) Output the sign of OR (OR 7) on each line of BIO7-0.
The micro-command "loop" which is used to iteratively alter the contents of the OR and SR registers has these modes:
(a) Decrement OR and store the result in OR.
(b) Decrement OR and store the result in OR and also store the result in one or both halves of SR. If enabled, SR3-0 receives the lower 4-bits of the decremented result while SR7-4, if enabled, receives the upper 4-bits of the decremented result.
(c) Transfer OR7-0 to SR7-4 and/or SR3-0.
(d) Increment OR7-4 and store in OR7-4, OR3-0 is unaffected.
(e) Increment OR7-4 and store in OR7-4 and SR7-4, OR3-0 and SR3-0 are unaffected.
(f) Compare OR7-4 to OR3-0 and indicate the result on the compare status outut pin. If the two half registers are not equal, increment OR7-4 and store in OR7-4. OR3-0 is unaffected.
(g) Same as 6 except the upper 4-bit result is stored in SR7-4 as well as OR7-4. SR3-0 and OR3-0 are unaffected.
(h) Compare OR7-4 and OR3-0 output the result on the compare status output. SR3-0 and OR3-0 are unaffected.
(i) Compare OR7-4 and OR3-0 as in (g) above, but transfer OR7-4 to SR7-4.
The loop status outputs from the RSU provide internal loop status information to the MCU. The "compare status output" is a logical 1 in the compare operations if OR7-4 =OR3-0. The loop borrow out signal is the borrow output from the loop decrement operations. The loop borrow output is true if OR7-0 is zero during a decrement operation or if OR7-4 is 1111 during a compare without increment operation.
FIGS. 5(a), (b), (c) and (d) illustrate a detailed circuit diagram of the logic circuits of the RSU unit. The operations of the circuits will be obvious to a person of ordinary skill in the art in light of the disclosure of these drawings, and a detailed discussion of the circuits will not be given.
Referring now to FIG. 7, a block diagram of a MCU unit is shown. The microcontroller (MCU) with micromemory has been designed to efficiently implement the entire control section of a computer. An extensive set of masking and data manipulation functions exist to provide for various combinations of external inputs to be mapped to the microaddress being generated. These types of functions are required to extract specific operation fields within macroinstructions, suboperation codes, and various combinations of status information. The MCU operations are determined by the microfield and synchronized by the clock. The MCU receives data from the bus input and the discrete I/O interface 128. The MCU generates a micromemory address output and certain discrete outputs.
The four stack registers (SR(3)-SR(0)) contained in stack register 116 allow linkages between various microsubroutines. The currently active stack register points to the next micromemory location to be executed. The MCU can go to a microsubroutine by pushing the stack and return from a subroutine by popping the stack. If too many levels of subroutine are called the stack wraps around and the oldest stack register is overwritten with the new stack value. A two bit Stack Pointer (SP) 117 keeps track of which SR is the currently used register.
The five operations Registers (R(4)-R(0)) contained in operations register 120 are dedicated to specific functions such as masking or saving common re-entry points. R(0) and R(1) are pointers to re-entry points that can be given control directly or conditionally. R(2) is a pair of 4-bit mapping registers that can be used to transfer execution to one of 16 micromemory locations depending on other conditions. R(3) is a maskable address register that can be loaded from microcode or from the external bus. R(4) is an address masking register.
The dual timer 118 is a pair of 16-bit counters that can be set up as inner and outer loop timers. Each timer is set up with a count and a set of branch conditions to be executed when the count runs out. Once started, a counter counts microcycles until terminated, pushed, or finished. The outer loop counter is suspended (pushed) by activiation of the inner loop counter. When the inner loop counter runs out, the outer loop counter starts counting on the count at which it was pushed.
The Address Generator 126 is used in determining the next address. The ability exists to mask and right justify (RJ) input fields, as needed, by implementing microcommand 3, the translate command; or microcommand 2, the map command. The control decode 122 of the MCU receives microcode input and based upon the bit structure, determines how the next address will be generated. It also determines which discretes and registers to use for the operations. The control decode section also indicates the proper stack register and governs the reading and writing of that register. It also starts, stops, resets, and restarts the dual timers. The Incrementer 124 receives code from the address generator and, in the normal mode, increments that address for the next microcode instruction. In the two's complement mode it acts as a counter and decrements the count.
Referring now to FIG. 8, the MCU has twelve (12) discrete inputs that connect various functions from other elements within the architecture that directly effect the flow of microcode. Examples of some of the signals that could be assigned to the discrete inputs are as follows: sign of the ALU output; overflow indication; carry out; indication of all zeros out of ALU; completion of load or store multiple; multiplier bits for sequential multiply operation; changing sign of dividend and sign of division for sequential divide operation; and interrupts.
A four bit register, loadable by four of the discrete inputs, is provided for machines requiring a condition code register (CCR). The MCU is the most efficient location for the CCR when discrete inputs are used for deriving status information. Means are also provided to load and store the contents of the CCR for exchange status requirements.
The translate command discussed below uses the discrete inputs to control the execution path. It offers the ability to select and move machine status information based on data from the GPU (or similar ALU) and other parts of the computer. Thus, status information is moved into an appropriate position for efficient micromemory address generation. The discrete inputs are organized as three (3) groups of four (4) bits each, and the CCR is considered a fourth group. The translate command selects one of the four groups and an immediate 4-bit mask identifies the specific bits within the group that are of interest. The masked bits are then right justified and merged into the appropriate low bits of the address pointer forming the next microaddress. The same structure is used by the conditional discrete setup command with the addition of a selected logical operation (AND, OR, XOR, XNOR) to be performed between the bits identified by the mask. A large selection of operand pair options are provided to minimize the number of set-up microinstructions required.
The discrete interface shown in FIG. 7 is further illustrated in FIG. 8. D(7)-D(0) are bidirectional signal pins while D(11)-D(8) are unidirectional inputs having a 4-bit register associated with them. The register bits indicate whether D(11)-D(8) are accepted in true or complement form. The ALU sign and overflow inputs to the controller are time multiplexed to contain the most significant shift input or output when required for the rotate function. Therefore, two pairs of discretes have the ability to pass data in either direction or complete the ALU circular shift macroinstruction.
The lower four discrete bits (D(3)-D(0)) are arranged for connecting the upper and lower shift bits of the GPUs (MXH(1)-(0)) and (MXL(1)-(0)). Under microcode control D(0) connects to D(2) and D(1) connects to D(3), or D(2) connects to D(0) and D(3) connects to D(1), or D(1) ex-or D(0) connects to D(2) and D(1) connects to D(3). The middle four discrete bits (D(7)-D(4)) are primarily used for storing and retrieving status bits. The Condition Code Register (CCR(3)-CCR(0)) can be output via D(7)-D(4) or can be loaded via D(7)-D(4). Alternatively, CCR(3)-CCR(0) can be loaded from (D(11)), D(10), D(1), (D(0)) or (D(11)), D(10), D(1), (D(1)-W-D(0)) or (D(11)), D(10), (D(1)-W-D(0), (DO).
Referring once again to FIG. 7, each MCU receives a 13-bit microcode control word from its associated micromemory unit. This control word is used to generate the next micromemory address. In addition to the microcode control word, each MCU receives an 8-bit data but input and several discrete signal inputs. The following Legend is a legend of the signals received by each MCU.
______________________________________LEGENDLABEL DESCRIPTION______________________________________MF0 Control Word Input - Bit0MF1 Control Word Input - Bit1MF2 Control Word Input - Bit2MF3 Control Word Input - Bit3MF4 Control Word Input - Bit4MF5 Control Word Input - Bit5MF6 Control Word Input - Bit6MF7 Control Word Input - Bit7MF8 Control Word Input - Bit8MF9 Control Word Input - Bit9MF10 Control Word Input - Bit10MF11 Control Word Input - Bit11MF12 Control Word Input - Bit12BI0 Bus Input - Bit0BI1 Bus Input - Bit1BI2 Bus Input - Bit2BI3 Bus Input - Bit3BI4 Bus Input - Bit4BI5 Bus Input - Bit5BI6 Bus Input - Bit6BI7 Bus Input - Bit7DI00 Discrete Input/Output - Bit0DI01 Discrete Input/Output - Bit1DI02 Discrete Input/Output - Bit2DI03 Discrete Input/Output - Bit3DI04 Discrete Input/Output - Bit4DI05 Discrete Input/Output - Bit5DI06 Discrete Input/Output - Bit6DI07 Discrete Input/Output - Bit7DI8 Discrete Input - Bit8DI9 Discrete Input - Bit9DI10 Discrete Input - Bit10DI11 Discrete Input - Bit11CLK Input ClockTSC Tri-State ControlRST Input ResetMA00 Memory Address Output - Bit0MA01 Memory Address Output - Bit1MA02 Memory Address Output - Bit2MA03 Memory Address Output - Bit3MA04 Memory Address Output - Bit4MA05 Memory Address Output - Bit5MA06 Memory Address Output - Bit6MA07 Memory Address Output - Bit7MA08 Memory Address Output - Bit8MA09 Memory Address Output - Bit9______________________________________
Each control word is decoded to generate one of eight possible MCU microcommands: (1) an unconditional branch immediate; (2) an unconditional branch immediate and link; (3) a map command; (4) a translate command; (5) a load registers command; (6) a conditional discrete operation; (7) a count command; and (8) a sub-operations command. The MCU control word is broken into 4 fields. The upper three bits (MF12-MF10) specify the microcommand. The next two bits (MF9-MF8) are page bits that are passed through the controller except during system reset, when they are forced to zero, and "same address", when previous value is used. (The page bits are used in systems having micromemory in excess of 256 words). The next 4-bit field (MF7-MF4) is either a branch address or a subcommand. The final 4-bit field (MF3-MF0) can be a branch address, a mask, or data to be loaded into a register half or counter half.
The following is a detailed discussion of each microcommand:
Microcommand 0 is the unconditional branch command. The lower 10 bits (MF9-MF0) are the address of the next microinstruction. The lower 8-bit portion of the microcommand is incremented and loaded into the currently active stack register (SR).
The unconditional branch is the normal mode for the transfer of program control to a predetermined address in micromemory, other than the next sequential address, with no interest in returning to the present location. This instruction is conventional in most controllers.
______________________________________UNCONDITIONAL BRANCH IMMEDIATE______________________________________ ##STR17##X = Immediate AddressOPERATION:X → MAO (Memory Address Out)(MAO) + 1 → SR (N) (Stack Register)______________________________________
Microcommand 1 is the branch and link command. The lower 10 bits are the next address as in command 0. However, the stack pointer is incremented prior to loading the incremented lower 8-bit portion of the microcommand into a stack register. The normal effect of a branch and link command is to leave the previously active SR pointing to the location in micromemory just past the branch and link command. Execution then progresses using the new SR until a return is executed. A return causes the controller to pick the next address from the previously active SR. BRL is the MCU method of providing for a subroutine "CALL". In addition to preforming an unconditional branch, BRL saves the return address (address of the BRL instruction+1) in its stack.
This instruction is used for the execution of frequently used sections of common code stored only once in the micromemory. An example of a subroutine entered in this manner would be one to perform normalization, which would be used by all floating point macroinstructions.
______________________________________UNCONDITIONAL BRANCH IMMEDIATE & LINK______________________________________ ##STR18##X = Immediate AddressOPERATION:X → MAO(MAO) + 1 → SR(N+1) (New StackRegister)______________________________________
Microcommand 2 is the map command. The lower 4-bits (MF3-MF0) are a mask and the second 4-bits (MF7-MF4) are sub-commands. The next micromemory address is determined as in Table 19. Any bits needed for the lower 4-bits of the microaddress that are not supplied by the masking operation are zeros. Table 20 shows an example of the masking operation for various mask values. The lower 8-bit portion of the address is incremented and loaded into the currently active SR.
The MAP instruction provides for making branches to one of several adjacent entry points based upon masked extraction from the contents of Registers R2 or R3, or the BUSIN data lines. Macroinstruction data is entered via the BUSIN port and contains a subset of execution codes (Add, Subtract, Multiply, etc.) of interest to the MCU. With the MAP instruction, the field of bits defining the subset is extracted and mapped to a predetermined section of micromemory. This technique is the most efficient means, from both space and speed considerations, of establishing a set of subroutine entry points.
______________________________________MAP______________________________________ ##STR19##P = Page bits (not required if Memory is 256 words or smaller)C = ControlM = Immediate MASKOPERATION:C: MF4 - Selection of BusIN (3-0)/R3 (3-0) for lower character sourceMF5 - Selection of Bus IN (7-4)/R3 (7-4) for upper character sourceMF6 - (OR) mask with upper character/ (AND) mask with lower characterMF7 - Both selected character out - masking operation specified by MF6/ character defined by MF6 as lower character source masked by m - upper character supplied by R2 (7-4)______________________________________
TABLE 19______________________________________COMMAND 2 SUBCOMMANDSM=MASK=MF3-MF0MICRO-COMMAND NEXT MICRO NEXT MICROMF7-MF4 ADDRESS BITS 7-4 ADDRESS BITS 3-0______________________________________0 0000 Bus In 7-4 VM Bus In 3-01 0001 Bus in 7-4 VM R3 3-02 0010 R3 7-4 VM Bus In 3-03 0011 R3 7-4 VM R3 3-04 0100 Bus In 7-4 (Bus 3-0 M) RJ5 0101 Bus In 7-4 (R3 3-0 M) RJ6 0110 R3 7-4 (Bus In 3-0 M) RJ7 0111 R3 7-4 (R3 3-0 M) RJ8 1000 R2 3-0 (Bus In 3-0 M) RJ9 1001 R2 3-0 (R3 3-0 M) RJA 1010 R2 3-0 (Bus In 7-4 M) RJB 1011 R2 3-0 (R3 7-4 M) RJC 1100 R2 7-4 (Bus In 3-0 M) RJD 1101 R2 7-4 (R3 3-0 M) RJE 1110 R2 7-4 (Bus In 7-4 M) RJF 1111 R2 7-4 (R3 7-4 M) RJ______________________________________ RJ = RIGHT JUSTIFY V = OR ∩ = AND-
TABLE 20______________________________________MASKING OPERATION RESULTSFOR MICROCOMMAND 2MASK BIT 4=0 BIT 4=1______________________________________0 0 01 000B0 (Bus In Bit 0) 000R302 000B1 000R313 00B1 B0 00R31 R304 000B2 000R325 00B2 B0 00R32 R306 00B2 B1 00R32 R317 0B2 B1 B0 0R32 R31 R308 000B3 000R339 00B3 B0 00R33 R30A 00B3 B1 00R33 R31B 0B3 B1 B0 0R33 R31 R30C 00B3 B2 00R33 R32D 0B3 B1 B0 0R33 R31 R30E 0B3 B2 B1 0R33 R32 R31F B3 B2 B1 B0 R33 R32 R31 R30______________________________________
Microcommand 3 is the translate command. Bits 3-0 form a mask and Bits 7-4 comprise a subcommand field. In the subcommands, Bits 5 and 4 select one of the 3 sets of discrete inputs or the conduction code register (CCR). Bits 7 and 6 select one of four masking operations the sixteen possible operations the sixteen possible operations are delineated in Table 21.
In Table 21, SRN is the current stack register and the notation "SRN 3-x" indicates that bits are brought down from SRN to fill the slots not filled by the masking operation. An example of the mask and right justify operation is given in Table 22.
In subcommands 4-8 and C-F, the stack register is not updated. This feature allows a routine such as multiply or divide to be set up in which the controller is continuously translating until interrupted by one of the timers running out. Otherwise, the lower 8-bit portion of the address is incremented and loaded into the currently active SR.
The XLT instruction is probably the second most common instruction, after BR. This instruction performs branches to addresses derived from various portions of the current SR, the discrete I/O lines, the condition code register (CCR), and a microcode supplied mask. Its most common use is as a "computed go to" based on the value of some combination of discrete lines. Virtually every test within the microcode will be performed with an XLT, whether it be testing to determine if indexing is to be performed or checking a shift count for completion. Some instructions (multiply, divide, count ones) use long sequences of XLT instructions, each translating to another XLT instruction, until finally some condition causes termination.
______________________________________TRANSLATE______________________________________ ##STR20##P = Page bitsC = ControlM = Immediate MASKOPERATION:P → MAO (9-8)C: MF4, MF5 - Input set selection (D (3-0)/D (7-4)/D (11-8)/CCR)MF6 - No Register Reference/Register ReferenceMR7 - If Register Reference, MASK/BIASAutomatic Mask Selection of Input set selectedInsertion of Discretes selected, right Justified, into SR(N).______________________________________
TABLE 21______________________________________TRANSLATE COMMAND SUBCOMMANDSNext Micro Address OutB7-4 MAO 7-4 MAO 3-0______________________________________0 0000 SRN7-4 V0000 SRN3-X, (D3-0 ∩M) RJ1 0001 SRN7-4 V0000 SRN3-x, (D7-4 ∩M) RJ2 0010 SRN7-4 V0000 SRN3-x, (D11-8 ∩M) RJ3 0011 SRN7-4 V0000 SRN3-x, (CCR3-0 ∩M) RJ4 0100 SRN7-4 V M SRN3-x, (D3-0 ∩R43-0) RJ5 0101 SRN7-4 V M SRN3-x, (D7-4 ∩R43-0) RJ6 0110 SRN7-4 V M SRN3-x, (D11-8 ∩R43-0) RJ7 0111 SRN7-4 V M SRN3-x, (CCR3-0 ∩R43-0) RJ8 1000 SRN7-4 V0001 SRN3-X, (D3-0 ∩M) RJ9 1001 SRN7-4 V0001 SRN3-x, (D7-4 ∩M) RJA 1010 SRN7-4 V0001 SRN3-x, (D11-8 ∩M) RJB 1011 SRN7-4 V0001 SRN3-x, (CCR3-0 ∩M) RJC 1100 SRN7-4 V R47-4 SRN3-x, (D3-0 ∩M) RJD 1101 SRN7-4 V R47-4 SRN3-x, (D7-4 ∩M) RJE 1110 SRN7-4 V R47-4 SRN3-x, (D11-8 ∩M) RJF 1111 SRN7-4 V R47-4 SRN3-x, (CCR3-0 ∩M)______________________________________ RJ The current Stack Register (SRN70) is updated for subops 0-3 and 8-B but is unchanged for subops 4-7 and C-F.
TABLE 22______________________________________M-CONTROLLER TRANSLATEINSTRUCTION LOWER BITSMASK RESULT______________________________________0 R3 R2 R1 R0 R3-0 is the current Stack1 R3 R2 R1 D0 Register (SRN)2 R3 R2 R1 D13 R3 R2 D1 D04 R3 R2 R1 D25 R3 R2 D2 D06 R3 R2 D2 D17 R3 D2 D1 D08 R3 R2 R3 D09 R3 R3 D1 D0A R3 R2 D3 D1B R3 D3 D1 D0C R3 R2 D3 D2D R3 D3 D2 D0E R3 D3 D2 D1F D3 D2 D1 D0______________________________________
Microcommand 4 is the load command. This instruction allows the lower 4-bits of the command MF3-MF0 to be loaded into one of 16 half registers. The next micromemory address comes from the current SR. The current SR is then incremented. Table 23 is a list of the destination addresses.
The load command is used to preset the value of the general registers (R0-R4), CCR, the TC mask register, or the counter holding registers. Since LD only loads 4 bits per instruction, two instructions are normally required for each register. With the exception of the CCR and R3 (temporary for BUSIN) the contents are only changed with a load instruction. The registers should primarily be used to store values required by frequently entered routines (e.g., fetch, operand classification, etc.) that would be loaded during the initialization sequence.
______________________________________LOAD______________________________________ ##STR21##P = Page BitsA = Register SelectionM = Immediate DataOPERATION:P, SR(N) → MAO9-0MAO (7-0) + → SR(N)M → (Register) A______________________________________
TABLE 23______________________________________LOAD COMMAND SUBCOMMANDSMICROCOMMAND DESTINATION OFMF7-MF4 MF3-MF0______________________________________0 R0 lower1 R0 upper2 R1 lower3 R1 upper4 R2 lower5 R2 upper6 R3 lower7 R3 upper8 R4 lower9 R4 upperA True/complement Mask for Discrete * 11-8B Condition Code Register (CCR3- CCR0)C Outer Timer Register lower holding register (TB3-TB0)D Outer Timer Register upper holding register (TB7-TB4)E Inner Timer Register lower holding register (TA3-TA0)F Inner Timer Register upper holding register (TA7-TA4)______________________________________ * 0 = ) True 1 = ) Complemented
Microcommand 5 is the conditional discrete setup command. MF5-MF4 select one of four sets of four signals from the discrete interface section to be examined. MF3-MF0 is a mask that selects which of the four signals selected by MF5-MF4 are to be examined by the controller. MF7-MF6 select the logical operation to be performed on the selected inputs. If the mask is zero, the current CCR is used as a mask. MF7-MF4 and the mask are latched for later use. Table 24 lists the microcommand 5 operations.
Once a command 5 has set up a discrete operation, the result of the prescribed logical function is sampled during subsequent operations until a command 7 is executed. During command 7, the MCU examines the latest discrete operation result, and branches based on a portion of the command 7 code. Unless the resulting branch directs the MCU to repeat the instruction (i.e. branch to the same address) the discrete operation is cancelled at the start of the next instruction. If no discrete operation is pending, then command 7 takes the "true" branch.
Due to data path conflicts, the discrete operation result is not sampled during command 3 or during most of the command 7 subcommands. The F subcommand of command 7 allows discrete result sampling and immediate branching on the sampled result.
The CD instruction is the most flexible of the conditional testing operations. CD allows for four logical operations (OR, AND, XOR, XNOR) to be performed on any combination of discrete lines within a set (D3-0, D7-4, or D11-8) or the CCR. The CD instruction does not itself initiate a branch operation, but rather establishes the conditions under which a branch will or will not later be taken. CD is most commonly used for such macro-level instructions as conditional jumps. Such macro-level instructions represent a large body of microcode, wherein each sequence is identical to the other except for the conditions to be tested (zero, less than zero, greater than or equal to zero, etc.). Using the CD instruction, a different set of conditional tests can be established (each in a single line of microcode) for each conditional jump. Once single sequence of microinstructions can then be executed to take the jump (or not), regardless of the conditions established for jumping.
______________________________________CONDITIONAL DISCRETE OPERATIONS______________________________________ ##STR22##P = Page Bits (for next address)C = Control & SelectionM = Immediate MaskOPERATION:C: MF5, MF4 - Input Set Selection (D(3-0)/D(7-4),/D(11-8)/CCR)MF7, MF6 - Operation on masked inputs (OR, AND, XOR, XNOR)P, SR(N) → MAO (9-0)MAO (7-0) + 1 → SR(N)______________________________________
TABLE 24______________________________________COMMAND 5 OPTIONS______________________________________ MF7-MF6 MF5-MF4______________________________________0 00 OR Discrete 3-01 01 AND Discrete 7-42 10 XOR Discrete 11-83 11 XNOR CCR 3-0______________________________________MASK RESULT IF MF7-MF4 = 000______________________________________0000 *0001 DI00010 DI10011 DI1 V DI00100 DI20101 DI2 V DI00110 DI2 V DI10111 DI2 V DI1 V DI01000 DI31001 DI3 V DI01010 DI3 V DI11011 DI3 V DI1 VDI01100 DI3 V DI21101 DI3 V DI2 V DI01110 DI3 V DI2 V DI11111 DI3 V DI2 V DI1 V DI0______________________________________ * Operation depends on the CCR when the command 5 was executed. DI = Discrete Interface
Microcommand 6 is the immediate count command. Command 6 has two subcommands determined by MF7. If MF7 is a zero, a sequential count is initiated; if MF7 is a one, an iteration count is initiated. In a sequential count subcommand, a count of 1 to 16, determined by MF3-MF0, is loaded into the outer counter unless the outer counter is active, in which case it is loaded into the inner counter and branch conditions determined by MF6-MF4 are loaded into the branch register for that counter. Execution then continues in normal sequence (i.e. starting at the location following the sequential count subcommand and continuing as prescribed by the subsequent instructions) until the count runs out. When the count equals zero a flag is set and the next instruction is determined by the branch instructions that were set up in the sequential count subcommand.
In an iteration count subcommand, the count determined by MF3-MF0 is also loaded into an available counter. However, the branch condition determined by MF6-MF4 is taken immediately. The instruction that results from the branch is then repeated until the count runs out. Execution then continues sequentially from the instruction that was repeated.
The protocol between the inner and outer counter is such that it is not possible for the inner counter to be active unless the outer loop counter has a count sequence pending. Attempting to initiate a third count while the inner counter is activated is an error which will cause the GPU controller to shut off both counters and take the branch conditions associated with the outer counter.
The command 6 branch options are illustrated in Table 25. In the R1 branch option, the contents of R1 provides the next memory address ((R1) MA out). The memory address out is then incremented and the result is loaded into the currently active stack register (MA out+1 SRn). In the MAP option, memory address bits 7 through 1 are provided by the currently active stack register. Memory address bit 0 is the exclusive-OR of discrete inputs D0 and D1. In the return option, the stack pointer is decremented (popped) and the memory address is provided with the content of the newly activated stack register. Execution continues using the newly activated stack register. In the link option, the stack is pushed by incrementing the stack pointer after having output a memory address from the currently active stack register. Link is used to go to a microsubroutine while RETURN is used to return from a microsubroutine. Finally, RETURN and LINK are used to return momentarily from a subroutine and go immediately back to the beginning of the subroutine, thus saving microcode when a subroutine must be executed multiple times.
The count/iterate instruction provides for two types of counting operations. Both types involve an initial count, which is counted down to zero, and a branch to a specified address, but differ as to the sequence in which these operations occur.
The count of the CNT instruction loads a counter and allows execution of microcode to continue at the next address. The specified branch is taken when the counter runs out, unless the count is terminated prematurely. Thus, the CNT instruction acts similar to "DO LOOPS" seen in some higher level languages, and can even be nested two deep. The CNT type of count is most useful in limiting sequences of microcode to a preset number of execution cycles, without consuming any time within the sequence to check for limits. Should the sequence terminate before the counter ends, the count can be forced to termination by the microcode. Specific uses for CNT involve the microcode sequences for such macro-level instructions as multiply and divide, where a known number of XLT instructions are to be executed before control is to be passed to a clean-up routine.
With the iterate or ITR instruction, the branch is taken immediately upon execution of the ITR. The instruction branched to is then executed "count" number of times, at which point execution continues in a normal fashion from the instruction which was repeated. A prime example of this involves the automatic repetition of a shift-one-bit instruction to make up a multibit shift.
______________________________________COUNT/ITERATE______________________________________ ##STR23##P = Page BitsC = ControlM = Immediate CountOPERATION:C: MF7 - Iteration (same address)/SequentialMF (4-6) - Address Sources (Same as True Condition Branch Options of Micro- command 7______________________________________
If Iterate, bits 4-6 specify the location of the instruction to be iterated. At the conclusion of the iteration, execution follows the normal path specified by the instruction that was iterated.
If sequential, the branch conditions specified by bits 4-6 are stored and a count is started. The next microlocation is the next micro-address. Execution continues in a normal fashion until the count runs out. The previously stored branch conditions then control the next address.
TABLE 25__________________________________________________________________________MICROCOMMAND 6BRANCH OPTIONSMF6-MF4 BRANCH CONDITION DESCRIPTION__________________________________________________________________________0 R1 Branch (R1) → MAout; MAout + 1 → SRn1 Next Address (SRn) → MAout; MAout + 1 → SRn2 MAP SRN (bits 7-1), DO DI → MAout, SRn unchanged3 Return (n) - 1 → n, (SRn) → MAout; MAout + 1 → SRn4 R1 Branch and (R1) → MAout, (n) + 1 → n; Link MAout + 1 → SRn5 R0 Branch (R0) → MAout; MAout + 1 → SRn6 Link (SRn) → MAout; (n) + 1 → n, MAout + 1 → SRn7 Return and Link (n) - 1 → n, SRn → MAout; (n) + 1 → n MAout + 1 → SRn__________________________________________________________________________ MAout = Memory Address out. SRn = Stack Register pointed to by the Stack Pointer, n. (R1) = The content of Register 1. XOR
Microcommand 7 is the sub-op and branch command. MF3-MF0 determine one of 16 discrete and timer operations. MF7-MF4 determine the next address.
The discrete and timer sub-operations are outlined in Table 26. Sub-ops 0-3 are different methods of loading the condition code register (CCR). Except for CCR1 in sub-op 2 the CCR is loaded with the selected information during clock phase 2. In sub-op 2, CCR1 is set during phase 2 if D1+D0 is a logical one; however, CCR1 is not reset if D1 t D10 is a logical zero.
Sub-op 8 causes the currently active counter to start over from the count in its holding register. If no counter is active, the outer loop counter is started. The branch conditions to be used by the counter are determined by MF7-MF4 as in Table 27. Unlike most of the other command 7 sub-ops, sub-op 8 causes the MCU to proceed to the following micromemory address. Should an active counter run out while a sub-op 8 is being executed the result is the same as attempting to initialize a third counter register, i.e. all counters are disabled and the outer loop branch conditions are taken.
Sub-op 9 initiates a sequential count as in command 6, except that the count comes from the holding register. The next address and count run-out results are as in sub-op 8, above. Sub-op C stops all counters and does not branch on the stored counter branch conditions. Sub-op D enables an iteration count as in command 6, except that the count comes from the holding register for the activated counter. Sub-op E cancels the currently active counter without taking the branch associated with that counter. Sub-op F samples the discrete operation set up by command 5.
Table 27 shows the branch conditions taken by sub-ops other than sub-ops 8 and 9 when no discrete operation is pending. The meaning of the branch conditions are the same as those in Table 25. Table 28 shows the branch conditions taken by sub-ops other than sub-ops 8 and 9 when a discrete operations is pending. The discrete operation is not sampled during a command 7 sub-ops O-E or a comman 3 (translate).
The sub-operations command is a double directive consisting of a data operation and a branch directive. The data operation is used to load discrete input lines into the condition code register, to transfer data between various discrete lines, and to manipulate the count register. The microcode sequences for all arithmetic macro-level instructions terminate with a sub-operations command specifying four of the discrete lines to be loaded into CCR, setting the carry, overflow, negative, and not-zero flags.
The second field is a branch directive specifying one of 16 non-unique branch possibilities. If there is a CD operation pending, each of these branch possibilities will have a true/false option, raising the total of 32 possible branches. It is among these sub branches that the return-from-subroutine is located. Several other branch modes allow indirect branching on the contents of a register. Virtually all macro-level instruction emulation sequences should utilize an indirect branch on Register 1 to branch to the "FETCH" sequence. This allows the branch and the loading of CCR to occur in a single instruction, rather than the two that a BR would require.
______________________________________SUB-OPERATIONS______________________________________ ##STR24##P = Page BitsC = Next AddressS = Suboperation______________________________________
TABLE 26______________________________________ COMMAND 7 SUB OPERATIONSMF 3-0 OPERATION______________________________________0 D11, D10, D1, D0 CCR3-CCR01 D11, D10, D1, D1 D0 CCR3-CCR02 D11, D10, (Di (Dc D0) V CCR2, D0 → CCR3-CCR0 (see Text)3 D7-D4 → CCR3-CCR04 Output CCR3-CCR0 → D7-D45 Output D1 D0 → D26 Output D3, D2 → D1, D07 Output D0→D38 Reload Sequential Count (see Text)9 Enable Sequential Count (see Text)A NO-OperationB Bus In 7-0 → R3 V = XORC Stop all counts V = ORD Enable Iteration Count (see Text)E Stop Current CountF Sample Discrete Operation (set up by command 5)______________________________________
TABLE 27______________________________________COMMAND 7 BRANCH CONDITIONS(NO DISCRETE OPERATION PENDING)MF 7-4 BRANCH______________________________________0 R1 Branch1 Next Address2 MAP3 Return4 R1 Branch and Link5 R0 Branch6 Link7 Return and Link8 R1 Branch9 Next AddressA Same AddressB ReturnC R1 BranchD R0 BranchE ReturnF Return and Link______________________________________ Exception: If sub op 8 or 9, ignore MF74 and take next address.
TABLE 28______________________________________COMMAND 7 BRANCH CONDITIONS(DISCRETE OPERATION PENDING)MF7-MF4 CONDITION TRUE CONDITION FALSE______________________________________0 R1 Branch Next Address1 Next Address Same Address2 MAP Next Address3 Return Next Address4 R1 Branch & Link Next Address5 R0 Branch Next Address6 Link Next Address7 Return & Link Next Address8 R1 Branch R0 Branch9 Next Address Return & LinkA Same Address Next AddressB Return R1 Branch & LinkC R1 Branch Same AddressD R0 Branch Return & LinkE Return Same AddressF Return & Link MAP______________________________________ Exception: If sub op 8 or 9, ignore MF74 and take next address. Attached hereto and marked as Exhibit A is a detailed circuit diagram of one implementation of the MCU unit discussed above, which is incorporated herein by reference for all purposes.
Turning now to FIG. 11, a block diagram of a typical bus interface unit (BIU), such as BIU 58, is shown. The primary functions provided by the BIU circuit are: (1) to provide a unidirectional interface to a bidirectional bus; (2) to provide a temporary storage for pipelining or ALU extension; (3) to provide protocol and priority for controlling bus activities; (4) to provide a path for transfers from one bus to another; and (5) to shift the temporarily stored data for floating point normalization or accumulator extension. FIGS. 12(a), (b), (c), (d) and (e) illustrate a detailed circuit diagram of a typical BIU circuit. It will be appreciated by one of ordinary skill in the art having the benefit of these drawings and the following disclosure how these circuits operate. Accordingly, a detailed discussion of the BIU circuit of FIG. 12 will not be given.
Still referring to FIG. 11, the BIU circuit is a dual 4-bit slice with one half connecting to the address bus and the other half connecting to the data bus. The uni-directional side of the transfer circuit is an integral part of the CPU data paths. The bi-directional side would be the common connection between the CPU, memory, I/O processor, etc. The source of output for each side of the transfer circuit is selectable from the opposite input of the same circuit, the register associated with each side (Register A 178 or Register B 174), or the selected input of the corresponding transfer circuit in the other 4-bit group (see FIG. 1, BIU 60, 62 64). The latter source allows the transfer of information from one bus to another.
The two sides of the BIU are identified as A and B. The sources of external data are the A and B bus unidirectional input (AI and BI) ports, and the A and B bus bidirectional (AIO and BIO) ports. The AIO and BIO ports both receive and transmit data using tri-state drivers. To complete the interface are two sets of transmit only ports which also incorporate tri-state drivers. These are the A and B bus unidirectional output (AO and BO).
Each of the two registers is identified with one 4-bit side i.e., Register A 178 and Register B 174. This means that data entering the BIU from either defined input (bidirectional or unidirectional) can be immediately loaded into its corresponding register only. When a register is selected as the source for an output, it will be the register associated with that output port. There are also means for transfer of data from one register to the other.
The shifter 182 in the BIU is used to shift the contents of either register. The input to the shifter 182 is selected to be either register A 178 or register B 174 and the destination is also selectable to be either register A or register B. Therefore, the contents of either register can be transferred to the other register via the shifter. There is no provision for exchange. The shifter has the ability to shift its input right 4, right 2, right 1, left 4, left 3, left 2, left 1, or no shift.
The protocol control lines are designed to establish controlled access to the bidirectional bus and eliminate the requirement for a bus scheduler in a less complex system of three or four interactive elements tied to the bus. A BIU circuit contains one set of control lines that are defined to be either send or receive by hardwiring one of the pins to VCC or ground. Therefore, a 16-bit bus interface using 4 BIU circuits will contain 4 sets of control lines that may be defined as required.
One set of control lines defined as "SEND" in one interface to the bus would be wired to and capable of establishing a link with another set of control lines defined as "RECEIVE" in another interface to the bus. These are permanent wiring and functional assignments within a given system. If information must be transferred in both directions between two interfaces on the bus, two sets of control lines are required--one set to establish the link in each direction.
A priority line serially connects all of the BIU circuits within an interface and resolves multiple requests of the same interface. The BIU at the top (most significant position, BIU 58 of FIG. 1) has the highest priority. In the sequence of establishing a link between two interfaces, the remaining sets of control modules in both interfaces are disabled from establishing any higher on lower priority links until the established link has been released.
The receive control (action to be taken by the receiving interface after a link has been established) is previously loaded and stored within a BIU circuit. The interface can therefore be set up by microcontrol to automatically store the information being received into the associated register under the interface link control. The information can then be taken from the register when required. This feature allows low overhead parallel operations required in high performance computers. Alternatively, information can be loaded into the registers and a send command given. The interface link and priority control will automatically complete the transfer when the bus is available.
One input pin of the BIU defines the control input to be either multiple format for microcontrol or single format for specific function control. In the single format mode the abilities do not exist to transfer data between the A and B side, to previously store the receive control word, or to use the shifter. The BIU circuits in the single mode can be set up as a reduced capability interface, such as that desirable for the memory, or can be used as a storage register tied to the CPU bus. In this latter mode, the BIU circuits are being used to implement the two status registers.
Referring still to FIG. 11, an 8-bit control word (CW7-0) selects various operations of the BIU. The most significant bit (CW7) determines whether the remaining bits have multiple definitions to utilize all the capability of the BIU or single definition for specific applications. CW7 is intended to be permanently wired as appropriate. The multiple format control word is intended for use under direct micro-control. The single format mode does not have all of the functions available but requires less active control inputs because some can be hardwired to describe the specific operation.
In single format, CW7 is a logical zero. The other bits are defined as follows:
(a) CW0 allows loading of register B 174 if CW0 =logic 1.
(b) CW1 allows loading of register A 178 if CW1 =logic 1.
(c) CW2 selects either AI or AIO as the input to the A section and BI or BIO as the input to the B section. Logic 1=bidirectional, logic 0=unidirectional.
(d) CW3 either register A 174 or AIO as the source for AO and either register A or AI as the source for AIO. Logic 1=circuit input, logic 0=register.
(e) CW4 selects AIO and BIO to be either inputs or outputs. Logic 0=bidirectional output enable.
(f) CW5 chooses either AIO only or both AIO and BIO to be outputs. Logic 0=output A part only, logic 1=output A and B parts.
(g) CW6 chooses either register B 174 or BIO as the source for BO and either register B 174 or BI as the source for BIO. Logic 1=circuit input, logic 0=register.
The multiple format control word, defined when CW7 is a logical one, selects one of four functions:
(a) Data Flow (DFLO).
(c) Load Register and Send Control (LDTR).
(d) Receive Control (LDRC).
The DFLO format defines the sources and flow direction of data. CW4 selects either AI and BI or AIO and BIO as the input data sources. CW3 and CW2 define which of AI, AIO, register A 178, BI or BIO will be output on AO or AIO. CW1 and CW0 control the BO or BIO outputs in the same way, selecting the sources to be BI, BIO, register B 174, AI or AIO.
The shift control format governs the function of the shifter. The contents of register A 178 can be shifted right one bit, right two bits, left one bit or left two bits. CW3-0 respectively define these functions. CW4 is not used.
The LDTR format commands are used to initiate a send and also to load register A 178 and register B 174. CW4 initiates a send. CW3 determines whether a read or write operation is being initialized. The other bits control the load operations. The possible operations are:
(a) Load register A 178, register B 174 or both and output on the respective IO.
(b) Output register A 178, register B 174 or both on the respective IO without loading.
(c) Pass data from AI to AIO, BI to BIO or both.
There is also a provision that if the microcode has initiated a send and the microcode is changed before the bus link is established, then register A 178 and register B 174 will default to be the sources for their respective outputs when the link is established.
The function of the LDRC format is to load a 3-bit receive control register (RC3-1) normally done during initialization. These bits in conjunction with the "operation input" determine the destination of data received by the BIU via established protocol link. Either C1 or C2 control can be used. The four functions selectable are:
(a) Load both register A 178 and register B 174 from AIO.
(b) Load register A 178 from AIO and register B 174 from BIO.
(c) Output AIO on AO and BIO on BO.
(d) Output AIO on both AO and BO.
CW4 is a bit used to initialize the protocol section, and CW0 is not used.
The bus protocol and control section establishes and controls BIU bus operation. The two directional lower and upper signals used to complete the shift paths between circuits are also used to unify the operation of all BIU circuits that make up the interface. The bus interface is disabled when the BIUs are being controlled by the shift format. Each BIU can set up the interface to either receive or send data. The corresponding status information is generated by the establishing BIU and received by the others. The four status states decoded from the two interconnecting signals are:
(a) C0: the wait state.
(b) C1: a receive state in which the BIU expects a response (as in a memory read)
(c) C2: a receive state in which no response is expected (as in a memory read).
(d) C3: the send state.
The BIU will always be in one of these states unless a shift is taking place. The shift functions use the same two interconnecting signals between BIUs.
Each BIU circuit contains eight protocol signals used to establish a link to transfer data to another like interface. One of the signals is hardwired (DIR) to program the remaining seven signals for sent protocol or receive protocol signals. The seven signals are defined as follows:
(a) A/RI (Acknowledge/Request In).
(b) R/DAI (Response/Data Available IN).
(c) RESET/OPIN (Reset/Operation IN).
(d) R/AO (Request/Acknowledge Out).
(e) OP/RO (Operation/Response Out).
(f) DAO (Data Available Out); used only for send.
(g) Clock In (same for both).
Each set of protocol signals would be wired to the corresponding set of protocol signals in another interface between which data is to be transferred via the bidirectional driver. If either interface is already in use, as determined by the state to be other than "CO", the protocol will hold until the interface is free and then establish the link. The linking protocol along with the receive control word automatically sets the flow direction, source, and destination of data for the respective interfaces.
Each BIU hs a set of protocol signals wired to either send or receive. Therefore, a 16-bit interface has four sets of protocol. Two sets of protocol at each interface are required to establish data transfers for both directions. A serial priority line and the status lines resolve simultaneous requests and prevent takeover of an on-going transfer. FIG. 13 illustrates a possible protocol connection scheme for a three interface bus system.
Referring now to FIG. 9, a block diagram of a typical ASU unit is shown. There are four primary functions provided by the ASU unit. First, each ASU unit provides resolution and synchronization of address inputs from two independent MCU units to a single section of micromemory 21. Second, each ASU unit provides storage of the current micromemory address while the next micromemory address is being generated in the MCUs. Third, each ASU unit operates to control the duration of the microcycle by controlling the duration of the microcycle clock signal outputted by the ASU. Fourth, each ASU unit provides clock offset for circuits requiring clock delay in relation to other circuits.
Each ASU contains the clock control and the microaddress control circuits for the emulator system. A 10-bit address can be received from two independent microcontrollers, MCU1 32 and MCU2 34 as A and B. The ASU selects the address and clock control bits from the A and B source according to the corresponding request bits and current status. There are three WAIT inputs which can stop the clock outputted by the ASU unit and two HOLD outputs that reflect the STATUS information to other ASUs.
The ASU selects the corresponding pair of bits to control the clock period. BP2 and BP1 are reflected by B selection, while AP2 and AP1 correspond to the A selection. For a clock of 8 MHz (125 nonosecond period) these bits control the clock cycle as shown in Table 29. Synchronized with the clock output is the enabling of the latch to hold the selected address (AO-9 or BO-9). The latches are disabled during phase one and are enabled during phase two of the clock period.
TABLE 29______________________________________ CLOCK OUTPUTAPI/BPI AP2/BP2 Phase 1 (0) Phase 2 (1)______________________________________0 0 125 ns 125 ns0 1 125 ns 250 ns1 0 250 ns 125 ns1 1 250 ns 250 ns______________________________________
Determination by the ASU as to which address period and inputs to select is controlled by the corresponding request inputs. Examples of control change based on Request sequences is illustrated in Table 30.
TABLE 30______________________________________CURRENT CYCLE NEXT CYCLE RESULTREQ. A REQ. B REQ. A REQ. B (Next Cycle)______________________________________0 0 0 0 no change0 0 0 1 B control0 0 1 0 A control0 0 1 1 A control0 1 0 0 B control0 1 0 1 B control0 1 1 0 A control0 1 1 1 B control1 0 0 0 A control1 0 0 1 B control1 0 1 0 A control1 0 1 1 A control______________________________________
In Table 30, "no change" means that whatever bit had control, will maintain control. In all cases, whenever a request occurs, it will not be honored until the end of the current cycle. Two HOLD outputs provide select STATUS information in the following manner. If the A/B request is received while a B/A request is being honored, then the AHOLD/BHOLD signal will be output until the A/B request is honored. This select status will remain until the B/A request has been dropped.
The three WAIT inputs are used to STOP the clock low. AWAIT received ruing phase one of the clock period will hold the clock low until the WAIT is released. A WAIT occurring in phase two of the clock period will allow the current cycle to finish and then hold the clock low. The address output is selected from the port (A or B) that had control when the WAIT occurred. After the WAIT is released, the current clock control and request bits determine how long the clock will remain in phase one.
There is an exception to the above cases that occurs when both WAIT 3 and REQUEST B are received. In this instance, the output clock is stopped as with any WAIT, and the B inputs are automatically selected. The B selection occurs even if the request A input is currently active. This function is used for immediate B-controller "takeover" without the A-side execution of an additional microinstruction.
The clock output of each ASU is normally used for the clock input of the same circuits being controlled by the section of micromemory the ASU is addressing. A delayed clock output is generated to circuits required to be stable until after the new microword is stable. The ROM ENABLE output is the NOR function of both clock signals and is used to skew the center transition of the cycle allowing more time for ROM access.
FIGS. 10(a) and (b) illustrate a detailed circuit diagram of an ASU unit. It will be appreciated by those of ordinary skill in the art having the benefit of the disclosures of these drawings and the specification how the circuits illustrated therein operate. Therefore, a detailed discussion will not be given.
Turning now to FIG. 14, in which a block diagram of the ICU 38 is shown, the ICU circuit contains an 8-level priority interrupt system and an 8-bit counter. Both functions are directly expandable with the additional circuits. Some of the features of the ICU are as follows: (1) incoming interrupt requests can either be strobed or captured as selected; (2) contains interrupt mask register, read and write via Bus I/O; (3) on chip storage of interrupt enable bit for multiple chip/class implementation; (4) internal enables and identification comparisons for common microcontrol in multiple ICU implementation; (5) contains 8-level concatenatable priority circuit with 3-bit encode of output; (6) storage of interrupt status while sampling vector; (7) separate input for counter clock; (8) counter can be read or written via Bus I/O; and (9) counter is directly concatenatable.
The ICU can be divided into four functional sections. The first to be described is the interrupt processing section which contains a mask register 214, vector generation, a circuit enable with internal storage, priority resolution, and interrupt request and acknowledge protocol circuitry. There are eight interrupt request lines (IRI 7-0) and eight corresponding acknowledge lines (IAO 7-0). Two inputs (IDI-0) are provided for chip identification.
The interrupt processing section can handle eight levels of interrupts with request 0 having the highest priority. When an interrupt request is received and its corresponding mask bit is true, all of the lower priority interrupts are inhibited. The interrupt system can be selected to operate in either a strobe or save mode. If the MODE line is true, interrupt requests are saved; if false, the device which initiates an interrupt request must hold it until the corresponding interrupt acknowledge is received. Interrupt requests may be initiated at any time. During an output vector instruction, the current interrupt status is latched and later incoming requests are isolated from the priority chain. The interrupt with highest priority is encoded into a three bit vector (V2-0). The interrupt acknowledge is then generated during phase two. This internally disables that incoming request from generating another interrupt until it has cycled. A 5-bit register 228, loadable from the bus is provided to extend the 3-bit interrupt vector to 8-bits.
The clear interrupt input will clear all pending interrupt requests being held and output all the Interrupt acknowledges. The interrupt acknowledge outputs can also be generated via an 8-bit input which comes from the bus as a control operation.
An internally stored interrupt system enable is provided for multiple circuit implementations requiring some circuits to be disabled. The enable or status bit is input and output via a separate bidirectional line. With the interrupt system disabled, the circuit will function like it is removed or shorted out. A priority input will be passed directly as the priority output and the circuit will not respond to a vector output instruction.
In a multiple ICU system with all circuits enabled, the last priority out signal indicates a pending interrupt. With the vector output control instruction requested of all the circuits, only those with an active interrupt request and no priority input (indicating a high priority interrupt) will respond by outputting the corresponding vector on the bidirectional bus.
Also included in the ICU circuits is an 8-bit synchronous up counter 210. The counter is intended for timer application and, therefore, has an independent clock input. The counter is also expandable by connecting the counter out of one circuit into the counter in of the next. The counter is loaded and read via the bidirectional bus with a control instruction input. Control instructions also enable and disable the counter.
The ICU incorporates a bidirectional bus interface. Inputs from the bus are loadable to the mask register 219, vector extension register 228, interrupt acknowledge outputs via request clear 230, and counter 210. Outputs from the ICU are derived from the mask register 214, interrupt vector 226 and 228 and counter 210.
A 6-bit control word (CB5-CB0) specifies which operation the ICU is to perform. CB5 and CB4 are compared to ID1 and ID0 (normally hardwired) respectively to select the circuit in which the operation denoted by CB3-0 is carried out. Table 31 is a list of control instructions and their corresponding operations. The clear interrupts instruction 2 overrides the MODE line so that pending interrupts that have been saved will be cleared and all interrupt acknowledge outputs will be cycled. Instruction 3 causes the state of the interrupt enable status register to be output.
The interrupt enable load instruction 4 loads the internal register from the bidirectional pin. Instruction 5 loads the mask register from the BIO. Instruction 6 outputs an interrupt acknowledge signal on all of the IAO lines for which the corresponding BIO bit is set. Instruction 7 loads the five-bit vector extension register (V7-3) from BIO bits 7-3. Each of these four operations is performed during phase two of the clock.
The instruction (8-B) output the designated contents onto the bidirectional bus. If the output vector instruction B is requested, the ICU output is enabled only if an interrupt request is pending, there is not a higher priority in, and the system is enabled. The control enable comparison (ID1, ID0 compared to CB5 and CB4) has no effect during instruction B.
The counter instructions C-F define operations to control the counter. A load instruction C will store BI0 7-0 into the counter register and hold until the next logic 0 to logic 1 transistion of the timer clock. It must be emphasized that the timer clock may or may not have any correlation with the clock which controls ICU control operations. An enable/disable instruction will start/stop the counter when the timer clock is low. If the instruction is received during the appropriate part of the time clock period, the action is taken immediately. If the instruction is received during the other half of the period, the action takes place after the next timer clock transition. A load and enable instruction loads the counter and enables it with the next logic 0 to logic 1 transition. Also implemented on the ICU circuit is a bidirectional switch and a tri-state driver.
TABLE 31______________________________________ICU INSTRUCTIONSCB3-0 OPERATION______________________________________0,1 No operation2 Clear Interrupts3 Output Status4 Load Status Register5 Load Mask Register from BIO6 Channel Acknowledge from BIO7 Load Vector Extension Register from BIO8,9 Output Counter Register to BIOA Output Mask Register to BIOB Output Vector to BIOC Disable CounterD Load & Disable CounterE Enable CounterF Load & Enable Counter______________________________________
FIGS. 15(a), (b) and (c) illustrate a detailed circuit diagram of an ICU unit. It will be appreciated by a person of ordinary skill in the art having the benefit of these drawings and the above detailed disclosure how the circuits of FIG. 15 operate. Accordingly, a detailed discussion of these circuits will not be given.
The following discussion describes the microcode, how the bits were distributed, and how the instructions are performed. Each microcode word consists of 112 bits divided into 7 banks of 16 bits each. These 7 banks are in turn divided into three sections of memory, one for each of the three address select units (ASU's). The left section (LM 20) is comprised of banks 1, 2 and 3, and is concerned chiefly with the arithmetic computations of UYK-20 macroinstructions. The right section (RM 24) is comprised of banks 5, 6 and 7, and is responsible for the address and addressing mode computation of the macroinstructions. The middle section of micromemory (MM 22) consists of bank 4 only, and is always slaved to either LM or RM. The micromemory breakdown by section and their associated bit assignments are shown in Table 32. MM 22 contains not only important control bits for LM's 20 GPU pairs and RSUs, but also the control field for the vital BIU, and is required by either side needing to communicate with macromemory. Either side of the emulator may request control of no banks, itself, itself and MM 22, or of all three sections as needed for computations. Further, execution by either side may be temporarily suspended pending availability of resources or data. Most noteably, instructions which require significant address computations prior to execution of the arithmetic function almost always suspend operation of LM 20 until the instruction operand has been obtained. This conserves micromemory, and simplifies arithmetic computations.
TABLE 32__________________________________________________________________________Bit Name Description Destination Bank__________________________________________________________________________96 MX8 MCU page LSB MCU1 195 MC2 MCU command MCU1 194 MC1 " MCU1 193 MC0 " MCU1 192 MX7 MCU operand MCU1 191 MX6 " MCU1 190 MX5 " MCU1 189 MX4 " MCU1 188 MX3 " MCU1 187 MX2 " MCU1 186 MX1 " MCU1 185 MX0 " MCU1 184 LNG 32 bit/long clock low GPU1,2 183 ECI Execute carry-in GPU1,2 182 DU2 Data type (upper) GPU2 181 DU1 " GPU2 180 DU0 " GPU2 279 DL2 Data type (lower) GPU1 278 DL1 " GPU1 277 DL0 " GPU1 276 MU2 Destination (upper) GPU2 275 ML2 Destination (lower) GPU1 274 M1 Destination (common) GPU1,2 273 M0 " GPU1,2 272 CA2 Connect code (M.S. GPU) GPU2 271 CA1 " GPU2 270 CA0 " GPU2 269 CB1 Connect code (CB2 = CD2 = 0) GPU1,2 268 CC2 " GPU1,2 267 CC1 " GPU1,2 266 CC0 " GPU1,2 265 CD1 Connect code (CB0 = CD0 = 1) GPU1,2 264 RQL Request left bank ASU1 (REQA) 363 A1 ALC control GPU1,2 362 A0 " GPU1,2 361 RQM Request middle bank ASU2 (REQB) 360 RQR Request right bank ASU3 (REQB) 359 DFU RSU default (upper) RSU2 358 DFL RSU default (lower) RSU1 357 MF8 RSU MC (combined) RSU1,2 356 MF7 RSU operand (combined) RSU1,2 355 MF6 " RSU1,2 354 MF5 " RSU1,2 353 MF4 " RSU1 352 MF3 " RSU1 351 MF2 " RSU1 350 MF1 " RSU1 349 MF0 " RSU1 348 MFIEU RSU MFIE (upper) RSU2 447 MFIEL RSU MFIE (lower) RSU1 446 S1 Source Select GPU1,2 445 S0 " GPU1,2 444 BIL RSU BIL (common) RSU1,2 443 LSHU RSU BSD (upper) RSU2 442 LSHL RSU BSD (lower) RSU1 441 LDB RSU BLD (common) RSU1,2 440 L35 Execute GPU tri-state GPU1 439 CW6 BIU command B1U1,2,3,4 438 CW5 " B1U1,2,3,4 437 CW4 " B1U1,2,3,4 436 CW3 " B1U1,2,3,4 435 CW2 " B1U1,2,3,4 434 CW1 " B1U1,2,3,4 433 CW0 " B1U1,2,3,4 432 R2 R select GPU3 531 R1 " GPU3 530 R0 " GPU3 529 CIA Carry-in GPU3 528 S1 Source Select GPU3 527 S0 " GPU3 526 D2 Data type GPU3 525 D0 " GPU3 524 A1 ALC control GPU3 523 M2 Destination GPU3 522 M1 " GPU3 521 M0 " GPU3 520 EXWT Execute wait ASU3 (WA173) 519 EQL Request left bank ASU1 (REQB) 518 RQM Request middle bank ASU2 (REQA) 517 RQR Request right bank ASU3 (REQA) 516 R3 R select (R3) GPU3 615 MC2 MCU command MCU3 614 MC1 " MCU3 613 MC0 " MCU3 612 MX7 MCU operand MCU3 611 MX6 " MCU3 610 MX5 " MCU3 69 MX4 " MCU3 68 MX3 " MCU3 67 MX2 " MCU3 66 MX1 " MCU3 65 MX0 " MCU3 64 MF3 RSU2 RSU2 63 MF2 " RSU2 62 MF1 " RSU2 61 MF0 " RSU2 6x16 T3 T select GPU3 7x15 T2 " GPU3 7x14 T1 " GPU3 7x13 T0 " GPU3 7x12 A0 ALC control GPU3 7__________________________________________________________________________
Execution of AN/UYK-20 emulation microcode begins at microlocation 000 (INIT) following power-up or RESET. Various hardware registers are initialized and the macroinstruction at location 0000 is fetched. Control is then transferred to microlocation 010 (FETCH) for instruction decoding. Several functions are performed at FETCH, some of which are necessary to decode the current instruction, others are useful in the later instruction execution.
(a) Load R3 of both MCUs with the respective outputs of the op-code transform ROM 36. This device maps the 6 bits of the "op-code" and the 2 bits of the "f" field into two addresses representing the instruction function (ADD, SUBTRACT, etc.) and the instruction addressing mode (RL, RR, RX, RI-2, RI-1, or RK).
(b) Load the OR and SR registers of both RSU's with the "a", "m" field of the macroinstruction.
(c) Preset the register select states in both RSUs.
(d) Transfer the contents of the UYK-20 register Rm to the RM holding register.
(e) Initiate a read for the contents of macromemory Pc+1.
(f) Branch to the address received from the op-code transform ROM (with control of MM 22 possessed by RM 24), or to the interrupt processing routine if an active interrupt is present.
Upon transfer from FETCH the RM 24 begins computation of the instruction operand, suspending or taking over control of the LM 20 as necessary. Once the operand is available the LM 20 is allowed to perform whatever operations were called for by the macroinstruction. Any necessary cleanup is performed, and control is returned to FETCH.
The LM-RM splitting of functions allows for a substantial savings in micromemory, while at the same time causing no reduction in macroinstruction execution speed. Because the LM 20 need only be concerned with the functional operation of the instruction, and not with the addressing mode, the same LM microcode can usually be used for all instructions performing the same function. For example, AR, AI, AK, and LA all perform the add function using the exact same microinstructions in LM 20. Further, AK, SUK, CK, LK, etc., all have their operand calculated by the same RM 20 microinstructions. This gives us an instruction matrix, with instruction function representing one ordinate, and addressing mode the other; and a minimum of total microcode. As a further bonus, where operand cleanup and instruction function computation do not require overlapping resources, both sides may operate simultaneously to improve macroinstruction throughput times.
The following Table 33 lists the contents of the micromemory 21 for emulating various macroinstruction of the AN/UYK-20.
TABLE 33__________________________________________________________________________AN/UYK-20 Emulation MicrocodeBank Number Left RightAddr. 1 2 3 4 5 6 7 Entry Entry__________________________________________________________________________00 0010 0C00 063F 4060 02D7 0010 0000 INIT01 4310 4E22 FF1E E070 EA50 4314 700002 4200 0C00 0600 0000 0057 4200 000003 4A80 0C00 0600 0000 0047 4A00 000004 06E0 0C00 0600 0015 0047 06E0 000005 7AA2 4C42 6718 E000 CBC7 BA90 E000 RXSUB RXSUB08 7AA2 4C42 6600 2000 ABC7 BA80 000009 7AA0 0C00 072A 4098 4757 3B40 60000A 7AA0 0C00 072F 4000 0047 00B0 00000B 7AA0 0C00 0600 0E92 0047 7130 00000C 63B2 4D22 DE00 0057 42CF 3B30 2800 MR0D 1C72 4E22 DC00 0046 0040 7AA0 00000E 00F2 4C22 DF23 C000 0950 06E9 00000F 7802 4042 6000 1013 0043 7AA0 0000 MLBFLY MLBFLY10 71B0 0C00 0600 0912 1257 71B0 0000 FETCH11 2302 4C22 872A C057 2603 2309 000014 0000 0000 0000 0098 475F 0080 100015 2300 0C00 8600 0098 4AD3 73A0 280016 0000 0000 0000 0098 4ADF 0080 280018 0000 0000 0000 0057 526F 00A0 200019 2300 0C00 8744 C000 0043 73A4 00001A 2300 0C00 8744 C000 0043 73A4 00001B 0000 0000 0000 0057 526F 00A0 20001C 0000 0000 0000 0000 C2DF 0090 20001D 7AA2 4C42 6600 0000 CAD7 0090 68001E 0000 0000 0000 0000 C2DF 0090 18001F 7AA3 FC42 6000 0000 CAD7 0090 680023 0000 0000 0000 0098 6A59 4D10 300024 3033 F126 0400 0080 5251 4CF0 200025 0C23 F222 9C00 0080 5250 71D0 200026 3033 F126 0400 0080 5251 71A0 200027 0C23 F222 9C00 0080 0047 0C20 000028 0290 0C00 9400 0098 6A51 0280 300029 7173 F122 6600 0028 0047 71A0 00002A 71A2 7D22 0600 2098 0047 71A0 00002B 7803 CD42 0600 2093 0047 0350 00002C 3930 0C00 9F14 0680 0040 70AE 0000 OP-402E 0000 0000 0000 2012 0040 0350 00002F 71B0 0C00 0000 0000 4257 78A0 0000 WJFETCH30 70A0 0C00 0000 0000 0047 70A0 000031 0280 0C00 9000 0000 0041 0400 000032 0852 CD22 E714 F098 6A53 0854 300033 0246 4C22 8714 E000 4853 0234 200034 0000 0000 0000 0000 0040 7800 0000 RR-SP (DUP)35 78A0 0C00 0000 00C5 6267 78A0 3000 FETCHHELD36 70A0 4C22 FA00 0000 2A50 78A0 1000 J40RR37 02F0 0C00 0000 0057 0267 02F0 100039 0000 0000 0000 0098 2B5B 3250 10003A 0000 0000 0000 0045 226B 03B0 10003B 0000 0000 0000 2012 0040 06E0 00003E 70A0 0C00 9800 0018 2A50 78A0 1000 J40RK3F 02F0 0C00 0000 0057 0267 02F0 100040 7802 DF42 6600 2013 0047 7AA0 0000 XRBF41 78A2 4022 6000 1013 0047 7AA0 0000 BUTFLY243 78A0 0C00 0611 0000 6047 78A0 2000 WAITFETCH44 0000 0000 0000 0098 6A5B 0450 300045 0000 0000 0000 0057 324B 0460 100046 0000 0000 0000 0000 224B 0470 100047 0000 0000 0000 0098 4A5B 0480 200048 0000 0000 0000 0057 23CB 0490 100049 0000 0000 0000 0000 004B 04A0 00004A 0000 0000 0000 0045 43CB 04B0 20004B 0000 0000 0000 209A 0040 0350 00004C 39F0 0C00 9F14 C000 0040 70AE 0000 OP-024D 5C13 FC22 9600 0080 0041 04E0 00004E 7003 FC22 9E00 0080 0040 70A0 00004F 7805 BD42 0600 0000 0047 70A0 000050 04D0 0C00 0600 0000 0047 04D0 000051 7003 FC42 0400 0000 4A57 0910 2000 RR52 7805 BD42 0600 0000 0040 70A0 0000 TCR53 7801 BD42 0600 0000 0040 70A0 0000 OCR54 7807 FD42 0600 0000 0040 70A0 0000 IROR55 0830 0C00 8602 C000 0043 0832 0000 IRTR56 70A0 0C00 0600 0000 0040 70A0 0000 ILLEGAL INSTRUCTION57 70A0 0C00 0600 0000 0040 70A0 000058 04D0 0C00 0600 0000 0047 0790 0000 NR59 70A0 0C00 0600 0000 0040 70A0 0000 ILL. INST.5A 780D BD42 0400 0000 0047 0A30 0000 TCDR5B 70A0 0C00 0600 0000 0040 70A0 0000 ILL. INST.5C 7800 9E42 0714 E000 0040 70A4 0000 DROR5D 0830 0C00 86FE C000 0043 083E 0000 DRTR5E 70A0 0C00 0600 0000 0040 70A0 00005F 70A0 0C00 0600 0000 0040 70A0 000060 0000 0000 0000 0000 004B 0610 0000 RK-DBL-SHIFT61 0000 0000 0000 0098 C7D3 32C0 700062 0000 0000 0000 0057 126B 0690 000063 0000 0000 0000 0057 126F 0430 000064 0000 0000 0000 E000 C7D3 32C0 7000 RL-DBL-SHIFT65 0000 0000 0000 0000 004B 0690 000067 0000 0000 0000 0000 004F 7A80 000068 0690 4C22 071C 0000 C7D3 32C0 7000 LRDR RR-DBL-SHIFT69 7A03 F262 6400 0000 D05B 5280 60006A 06B0 4C22 071C 0000 D253 71F0 6000 ASDR6B 7A03 F282 6400 0000 004F 0A30 00006C 06D0 4C22 071C E000 C7D3 32C0 7000 ALDR RL-SNGL-SH6D 7A23 F123 6400 0000 004B 0710 00006E 06F0 4C22 071C 0057 1257 0430 0000 CLDR FTCHIN6F 7A73 F122 6400 0000 004F 7A80 000070 0710 4C22 071C 0000 C7D3 32C0 7000 LLRS-LRS-LRSR RR-SNGL-SHIFT71 7A03 F266 6600 0000 D05B 5280 600072 0730 4C22 071C 0000 D253 71F0 6000 ARS-LARS-ARSR73 7A03 F288 6600 0000 004F 78A0 000074 0750 4C22 071C 0000 004B 0750 0000 ALS-LALS-ALSR RK-SNGL-SHIFT75 7A23 F133 6600 0098 C7D3 32C0 700076 0770 4C22 071C 0057 126B 0710 0000 CLS-LCLS-CLSR77 7A73 F122 6600 0057 126F 0430 000078 0000 0000 0000 0000 004B 0A60 0000 RI2-STORE-DBL79 5013 FC42 9600 0080 0040 04E0 00007B 7802 4E42 0600 2012 0040 0350 00007C 07D0 0C00 0124 0000 004B 07D0 0000 LXI -- LXI7D 07B7 FD22 0000 0098 6A5B 07E0 20007E 07F3 FC22 E74D C0D7 2240 07F4 1000 SDI7F 0AA0 0C00 8746 0080 3250 07B0 100080 7802 4E42 671C 0000 004B 0810 0000 LL-LR-LI-LK RI281 0000 0000 0000 0098 6A5B 0AA0 200082 7802 DE42 071C 0000 0041 73A0 0000 LA-AR-AI-AK83 7802 DE42 071E E000 0040 70A5 000084 7804 9E42 071C 0000 0040 7800 0000 SUK-LSU-SUR-SUIRR-SNGL-MATH85 08A3 FC22 9714 0046 0041 08A4 000086 0872 4C22 F71C 0000 0041 69D0 0000 LC-CR-CI-CK IT1587 7804 9C42 B600 0000 0041 0820 00008A 3040 CE22 6600 3028 0041 1400 00008B 3047 CE22 0600 3028 0047 0350 00008C 0400 EE22 DE00 0000 004B 08D0 1000 XORK-XORR-XORIRI2-SNGL-MATH8D 0000 0000 0000 0098 6A5B 08E0 30008E 0000 0000 0000 00D7 224B 08F0 10008F 0000 0000 0000 0000 004B 02E0 000090 7802 DE42 4600 0000 0000 0000 0000 ANDK-ANDR-ANDI91 0000 0000 0000 0000 424F 5C10 200092 7802 DE42 6600 0000 004F 70A0 0000 ORK-ORI-ORR93 7807 FD42 0600 0000 0047 70A0 000094 780C 9C42 0600 0000 004B 0390 0000 CDR-CDI RK-SNGL-MATH9C 790C 9E42 9C00 0000 004B 0440 0000 SUDI-SUDR RI2-DBL-MATH9D 0412 4C22 9723 D000 0040 7AA9 0000 BUTERFLY9E 790A DE42 9C00 0000 004B 0350 0000 ADR-ADI9F 0412 4C22 9F23 D000 0040 7AA9 0000A0 0AE2 4C22 C714 E000 0040 78A4 0000 LAD RL-SNGL-MATHA1 710C 9D42 851C E098 0040 0A24 0000A2 0A73 CF22 E600 2093 0047 78A0 0000A3 0412 4C22 9F23 D000 0040 0109 0000 BTFLYA4 0A50 0C00 0000 0000 0047 48F0 0000 OP-40RK RK40JMPA5 03D0 0C00 011E 4080 0047 35E0 0000A6 78A2 4C22 4714 0098 6A5B 0A70 3000 SZIA7 78A2 7F22 3000 C0DF 23C0 0AED 1000A8 0A12 4C22 C714 E014 0040 0A14 0000 LSUD RL-DBL-MATHAA 78A3 FD22 6714 E05F 0240 09E4 1000 STORE STOREAB 0352 4E22 0600 0000 004B 0350 0000AC 0AD0 0C00 0000 0000 0047 48F0 0000 OP-40-RR-JMP RR40JMPAD 0350 0C00 011E C080 0047 35EE 0000AE 7102 DD42 851C C000 3253 0AA6 1000AF 0A73 CF22 E600 0000 0000 0000 0000B0 5283 FC52 0600 0080 0047 0B10 0000 JZ-JXR RR44-47JMPB1 0363 FC52 0600 0080 0047 0360 0000B2 5213 FC52 0600 0000 0000 0000 0000 JP-JPRB4 5E83 FC52 0600 0080 0047 0B50 0000 JNZ-JNZR RK44-47JMPB5 0360 FC52 0600 0080 0047 03E0 0000B6 5E13 FC52 0600 0000 0000 0000 0000 JN-JNRB8 7802 DE42 6718 E000 0043 78A0 0000 SBRB9 70A3 FC42 6600 0000 C5D0 F0A0 D800BA 7800 9E42 4718 0000 0000 0000 0000 CBRBB 0AB0 0C00 8746 C012 0047 03F6 0000BC 7802 DC42 4718 0000 004F 1050 0000 TBR RX-SPMATHBD 0000 0000 0000 0057 42CB 0BE0 2800BE 790A 4D42 FC00 0000 004B 03B0 0000 LDIBF 0412 4C22 9723 D000 0040 0359 0000C0 0C13 FC22 FF24 0000 0000 0000 0000 SXIC1 0357 FD22 9E00 0000 0040 0350 0000C2 0C33 CD22 FF4D E093 0040 71CD 0000C3 0BB3 CF22 F94D E045 0040 0BBD 0000C4 39F0 0C00 9F1E C000 0040 010E 0000 OP-03 OP-03C5 7AA0 0C00 0655 C000 42A0 0C65 C800C6 0B92 4C42 6714 E000 A7D0 8B9E 2000__________________________________________________________________________
Referring once again to FIG. 1, op-code transform ROM 36 operates to strip the operand address information and the instruction op-code from each macroinstruction. On the basis of this information, ROM 36 generates two 8-bit addresses, one to each MCU unit. An 8-bit address obtained from the 6-bit op-code followed by the 2-bit "f" field of the UYK-20 macroinstructions is used to address ROM 36. Each address location outputs the following 12-bit address field that is used to generate the 8-bit addresses to the MCUs. ##STR25## The address for the left MCU 32 is comprised of (L7 . . . L1)*2, while the address for the right MCU 34 is comprised of (L7, R6 . . . R2)*4.
The following Table 34 is a listing of the contents of ROM 36 for selected ones of the UYK-20 macroinstructions. While all of the macroinstructions to emulate the entire instruction set for the UYK-20 is not given in Table 34. those given, when taken in conjunction with the microcode contained in micromemory 21 and listed in Table 33, should serve as an illustrative example of how the ROM 36 is programmed.
TABLE 34______________________________________Op Valu Instruction Op Valu Instruction______________________________________15 7DF LXI 52 865 CK18 BAE CBR 53 86F C1C BCE TBR 58 0CD MR20 71C LRSR 59 F43 MI22 71D LRS 5A CC5 MK24 73C ARSR 5B CCF M25 AA0 SI 60 901 ANDR26 73D ARS 61 903 ANDI28 69A LRDR 62 905 ANDK 63 90F AND2A 698 LRD 64 921 ORR2C 6BA ARDR 65 923 ORI2E 6B8 ARD 66 925 ORK30 75C ALSR 67 92F OR32 75D ALS 68 8C1 XORR34 77C CLSR 69 8C3 XORI36 77D CLS 6A 8C5 XORK38 6DA ALDR 6B 8CF XOR3A 6D8 ALD 80 ACB J4OR3C 6FA CLDR 94 8C5 JNZR3D A60 SZI C0 61B LRSL3E 6F8 CLD C1 73B ARSL40 841 SUR C2 699 LRDL41 843 SUI C3 6B9 ARDL42 845 SUK C4 75B ALSL43 84F SU44 9C1 SUDR C5 77B CLSL45 9C7 SUDI C6 6D9 ALDL48 821 AR C7 6F9 CLDL49 823 AI C8 848 SUL4A 825 AK C9 A88 SUDL4B 82F A CA 828 AL4C 9E1 ADR CB A08 ADL4D 9E7 ADI CC 808 LL50 861 CR CD 868 CL51 863 CI CE F08 LMUL______________________________________
Table 35 illustrates a typical sequence in micromemory addresses and ASU clock signals during the emulation of a particular macroinstruction of the AN/UYK-20 computer. The macromemory address and the contents of the addressed location (the macroinstruction) is illustrated in Table 35 opposite the LM 20 and the RM 24 addresses; the logic levels of the clock signals outputted by each ASU unit (LC, MC, RC); and the ownership of each of the three ASU units, i.e., which MCU, the right R or the left L, has control of the ASU and is specifying the micromemory address.
The following discussion refers to the corresponding number reference identified in Table 35 and identifies the functions which are being performed in those microcycle.
(1) The UYK-20 instruction (fetched previously) is LK R2, '41 'X (load next word into R2). Cleanup from previous instruction is performed. Fetch (read) of next UYK-20 word is initiated. Right MCU in total control.
(2) The current instruction op-code and mode field for LK (0616) is passed through op-code transform ROM 36. Each MCU controls his respective side. Middle is controlled by right MCU 34.
(3) Each MCU goes to perform its respective function: the left to perform a "load", the right to perform an Rk -mode operand derivation. Neither side is aware of what the other will be doing. However, the right MCU 34 knows that execution of the instruction (whatever it may be) cannot proceed until operand derivation is complete. Therefore, the right MCU 34 forces the left MCU 32 (and entire left side of the machine) to go to sleep, i.e., no LC is generated from ASU 26.
(4) The right MCU 34 checks for possible register indexing. Note that the left ASU's 32 clock does not change (i.e. asleep).
(5) Operand is available. Left side is allowed to awaken and complete load operation (clock change shows up at next instruction).
(6) The left MCU, awake and having completed the load, attempts to return to FETCH (address 10). The right MCU 34 however, realizes that another read of macromemory is necessary since the last read produced data rather than another instruction.
(7) The right MCU takes control of the entire machine to await completion of the read.
(8) The right MCU 34 forces all three parts of the machine back through FETCH to process th new instruction (another "load", but indexed; LK R3, `33` X, R2).
(9) Same as (4), except register indexing is specified. Address 3A adds contents of R2 to operand.
TABLE 35__________________________________________________________________________ ASU MACRO MACRO OWNERSHIP MEMORY MEMORY LAD RAD LC MC RC L M R ADDRESS DATA COMMENT__________________________________________________________________________ 010 010 111 LLR 0000 0620 LK Instruction 010 010 000 LLR 0000 0620 Execution Begins 011 011 111 RRR 0000 0620 011 011 000 RRR 0000 0620 206 206 111 LRR 0001 0041 Fetch of Next Word 206 206 000 LRR 0001 0041 080 094 111 LRR 0001 0041 080 094 000 LRR 0001 0041 080 039 011 LRR 0001 0041 Not Indexed 080 039 000 LRR 0001 0041 080 03B 011 LRR 0001 0041 Data Stored in Register 080 03B 000 LRR 0001 0041 2 010 06E 111 LRR 0001 0041 010 06E 000 LRR 0001 0041 043 043 111 RRR 0002 0632 Next Instruction Fetched 043 043 000 RRR 0002 0632 010 010 111 RRR 0002 0632 LK Instruction 010 010 000 RRR 0002 0632 Execution Begins 011 011 111 RRR 0002 0632 011 011 000 RRR 0002 0632 206 206 111 LRR 0003 0033 Fetch of Next Word 206 206 000 LRR 0003 0033 080 094 111 LRR 0003 0033 080 094 000 LRR 0003 0033 080 039 011 LRR 0003 0033 Indexed 080 039 000 LRR 0003 0033 080 03A 011 LRR 0003 0033 Index Added to Operand 080 03A 000 LRR 0003 0033 080 03B 011 LRR 0003 0033 Operand Plus Index 080 03B 000 LRR 0003 0033 Stored in Register 2 010 06E 111 LRR 0003 0033 010 06E 000 LRR 0003 0033 043 043 111 RRR 0004 0523 043 043 000 RRR 0004 0523 010 010 111 RRR 0004 0523__________________________________________________________________________ LEGEND LAD = Left Micromemory Address RAD = Right Micromemory Address LC = Left Clock from Left ASU MC = Middle Clock from Middle ASU RC = Right Clock from Right ASUASU Ownership (L=left, M=middle, R=right) = Microcontroller specifying address out of ASU (L=left microcontroller, R=right microcontroller)__________________________________________________________________________
In describing the invention, reference has been made to a preferred embodiment. However, those skilled in the art and familiar with the disclosure of the invention may recognize additions, deletions, substitutions or other modifications which would fall within the purview of the invention as defined in the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3544969 *||Nov 27, 1967||Dec 1, 1970||Standard Computer Corp||Language independent computer|
|US3651482 *||Apr 3, 1968||Mar 21, 1972||Honeywell Inc||Interlocking data subprocessors|
|US3760365 *||Dec 30, 1971||Sep 18, 1973||Ibm||Multiprocessing computing system with task assignment at the instruction level|
|US3766532 *||Apr 28, 1972||Oct 16, 1973||Nanodata Corp||Data processing system having two levels of program control|
|US3766533 *||May 30, 1972||Oct 16, 1973||Ibm||Processor utilizing one holographic array and a plurality of photoresponsive storage arrays for high paging performance|
|US3955180 *||Jan 2, 1974||May 4, 1976||Honeywell Information Systems Inc.||Table driven emulation system|
|US4041461 *||Jul 25, 1975||Aug 9, 1977||International Business Machines Corporation||Signal analyzer system|
|US4096467 *||Dec 15, 1976||Jun 20, 1978||E M I Limited||Ultrasonic image conversion|
|US4156903 *||Feb 28, 1974||May 29, 1979||Burroughs Corporation||Data driven digital data processor|
|US4167781 *||Oct 12, 1976||Sep 11, 1979||Fairchild Camera And Instrument Corporation||Microprocessor system having a single central processing unit shared by a plurality of subsystems each having a memory|
|US4199811 *||Sep 2, 1977||Apr 22, 1980||Sperry Corporation||Microprogrammable computer utilizing concurrently operating processors|
|US4266281 *||Apr 2, 1979||May 5, 1981||Allen-Bradley Company||Microprogrammed programmable controller|
|US4313161 *||Nov 13, 1979||Jan 26, 1982||International Business Machines Corporation||Shared storage for multiple processor systems|
|US4325116 *||Aug 21, 1979||Apr 13, 1982||International Business Machines Corporation||Parallel storage access by multiprocessors|
|1||*||N. Tinkelpaugh & R. E. Conklin, A Radiation Hardened Microprocessor Circuit Set and Its Application, May, 1979.|
|2||*||R. E. Fosdick, Technical Proposal for CMOS/SOS Standard Computer Emulation, Apr. 14, 1978.|
|3||*||R. Fosdick, CMOS/SOS Standard Computer Emulation, Sep., 1978.|
|4||*||R. Fosdick, CMOS/SOS Standard Computer Simulation AN/UKY-20 Emulator, Design Report-8/79, Final Design Report-2/80.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US4467415 *||Aug 31, 1981||Aug 21, 1984||Nippon Electric Co., Ltd.||High-speed microprogram control apparatus with decreased control storage requirements|
|US4521858 *||Mar 11, 1983||Jun 4, 1985||Technology Marketing, Inc.||Flexible addressing and sequencing system for operand memory and control store using dedicated micro-address registers loaded solely from alu|
|US4533992 *||Feb 22, 1982||Aug 6, 1985||Texas Instruments Incorporated||Microcomputer having shifter in ALU input|
|US4633417 *||Jun 20, 1984||Dec 30, 1986||Step Engineering||Emulator for non-fixed instruction set VLSI devices|
|US4638423 *||Mar 6, 1985||Jan 20, 1987||Motorola, Inc.||Emulating computer|
|US4654788 *||Jun 15, 1983||Mar 31, 1987||Honeywell Information Systems Inc.||Asynchronous multiport parallel access memory system for use in a single board computer system|
|US4685080 *||Feb 22, 1982||Aug 4, 1987||International Business Machines Corp.||Microword generation mechanism utilizing separate programmable logic arrays for first and second microwords|
|US4727480 *||Jul 9, 1984||Feb 23, 1988||Wang Laboratories, Inc.||Emulation of a data processing system|
|US4736317 *||Jul 17, 1985||Apr 5, 1988||Syracuse University||Microprogram-coupled multiple-microprocessor module with 32-bit byte width formed of 8-bit byte width microprocessors|
|US4736338 *||Oct 7, 1986||Apr 5, 1988||Silicon Solutions Corporation||Programmable look up system|
|US4761755 *||Jul 11, 1984||Aug 2, 1988||Prime Computer, Inc.||Data processing system and method having an improved arithmetic unit|
|US4794527 *||Jan 29, 1986||Dec 27, 1988||Digital Equipment Corporation||Microprogrammed data processing system using latch circuits to access different control stores with the same instruction at different times|
|US4812971 *||Jun 16, 1988||Mar 14, 1989||Digital Equipment Corporation||Central processing unit for a digital computer|
|US4812975 *||Jul 11, 1984||Mar 14, 1989||Hitachi, Ltd.||Emulation method|
|US4833594 *||Dec 22, 1986||May 23, 1989||International Business Machines||Method of tailoring an operating system|
|US4873628 *||Apr 18, 1988||Oct 10, 1989||Sharp Kabushiki Kaisha||Command translating computer|
|US4893235 *||Jun 16, 1988||Jan 9, 1990||Digital Equipment Corporation||Central processing unit for a digital computer|
|US4931989 *||Apr 15, 1987||Jun 5, 1990||International Business Machines Corporation||Microword generation mechanism utilizing a separate programmable logic array for first microwords|
|US4951195 *||Feb 1, 1988||Aug 21, 1990||International Business Machines Corporation||Condition code graph analysis for simulating a CPU processor|
|US4972317 *||Aug 4, 1989||Nov 20, 1990||International Business Machines Corp.||Microprocessor implemented data processing system capable of emulating execution of special instructions not within the established microprocessor instruction set by switching access from a main store portion of a memory|
|US5062034 *||Nov 20, 1989||Oct 29, 1991||U.S. Philips Corporation||Device for emulating a microcontroller using a parent bond-out microcontroller and a derivative non-bond-out microcontroller|
|US5072372 *||Mar 3, 1989||Dec 10, 1991||Sanders Associates||Indirect literal expansion for computer instruction sets|
|US5095426 *||Jun 6, 1989||Mar 10, 1992||Nec Corporation||Data processing system for effectively handling exceptions during execution of two different types of instructions|
|US5101342 *||May 16, 1989||Mar 31, 1992||Kabushiki Kaisha Toshiba||Multiple processor data processing system with processors of varying kinds|
|US5109497 *||Jan 27, 1989||Apr 28, 1992||Hughes Aircraft Company||Arithmetic element controller for controlling data, control and micro store memories|
|US5129064 *||Aug 27, 1990||Jul 7, 1992||International Business Machines Corporation||System and method for simulating the I/O of a processing system|
|US5146606 *||Mar 21, 1990||Sep 8, 1992||Digital Equipment Corporation||Systems for interconnecting and configuring plurality of memory elements by control of mode signals|
|US5148528 *||Feb 3, 1989||Sep 15, 1992||Digital Equipment Corporation||Method and apparatus for simultaneously decoding three operands in a variable length instruction when one of the operands is also of variable length|
|US5159674 *||May 17, 1990||Oct 27, 1992||Siemens Aktiengesellschaft||Method for supplying microcommands to multiple independent functional units having a next microcommand available during execution of a current microcommand|
|US5167023 *||Dec 7, 1990||Nov 24, 1992||International Business Machines||Translating a dynamic transfer control instruction address in a simulated CPU processor|
|US5168570 *||Jun 11, 1990||Dec 1, 1992||Supercomputer Systems Limited Partnership||Method and apparatus for a multiple request toggling priority system|
|US5170484 *||Mar 21, 1990||Dec 8, 1992||Digital Equipment Corporation||Massively parallel array processing system|
|US5179696 *||Jun 24, 1991||Jan 12, 1993||Nec Corporation||Generator detecting internal and external ready signals for generating a bus cycle end signal for microprocessor debugging operation|
|US5202975 *||Jun 10, 1992||Apr 13, 1993||Supercomputer Systems Limited Partnership||Method for optimizing instruction scheduling for a processor having multiple functional resources|
|US5202976 *||Jun 21, 1991||Apr 13, 1993||Hewlett-Packard Company||Method and apparatus for coordinating measurement activity upon a plurality of emulators|
|US5222230 *||Nov 20, 1989||Jun 22, 1993||Texas Instruments Incorporated||Circuitry for transferring data from a data bus and temporary register into a plurality of input registers on clock edges|
|US5230079 *||Nov 16, 1987||Jul 20, 1993||Digital Equipment Corporation||Massively parallel array processing system with processors selectively accessing memory module locations using address in microword or in address register|
|US5301302 *||Apr 2, 1991||Apr 5, 1994||International Business Machines Corporation||Memory mapping and special write detection in a system and method for simulating a CPU processor|
|US5305462 *||Jun 3, 1993||Apr 19, 1994||Digital Equipment Corporation||Mechanism for broadcasting data in a massively parallell array processing system|
|US5333287 *||Feb 23, 1993||Jul 26, 1994||International Business Machines Corporation||System for executing microinstruction routines by using hardware to calculate initialization parameters required therefore based upon processor status and control parameters|
|US5423012 *||Nov 18, 1992||Jun 6, 1995||Mitsubishi Denki Kabushiki Kaisha||Apparatus and method in a computer for executing calculation instructions and data instructions having uniform word lengths|
|US5454088 *||May 17, 1993||Sep 26, 1995||Mitsubishi Denki Kabushiki Kaisha||Microprogram control device for controlling data path section including designation of instruction cycle values|
|US5495617 *||Jun 2, 1995||Feb 27, 1996||Nec Corporation||On demand powering of necesssary portions of execution unit by decoding instruction word field indications which unit is required for execution|
|US5724590 *||Sep 18, 1996||Mar 3, 1998||Lucent Technologies Inc.||Technique for executing translated software|
|US5784638 *||Feb 22, 1996||Jul 21, 1998||International Business Machines Corporation||Computer system supporting control transfers between two architectures|
|US5918029 *||Sep 8, 1997||Jun 29, 1999||Digital Equipment Corporation||Bus interface slicing mechanism allowing for a control/data-path slice|
|US6108763 *||Oct 3, 1994||Aug 22, 2000||Grondalski; Robert S.||Simultaneous parity generating/reading circuit for massively parallel processing systems|
|US6108765||Oct 8, 1997||Aug 22, 2000||Texas Instruments Incorporated||Device for digital signal processing|
|US6360194 *||Sep 8, 1998||Mar 19, 2002||Bull Hn Information Systems Inc.||Different word size multiprocessor emulation|
|US6650317||Jan 5, 1995||Nov 18, 2003||Texas Instruments Incorporated||Variable function programmed calculator|
|US7213131 *||Jan 15, 2004||May 1, 2007||Microunity Systems Engineering, Inc.||Programmable processor and method for partitioned group element selection operation|
|US7260515 *||Mar 28, 2002||Aug 21, 2007||Sun Microsystems, Inc.||Method and apparatus for simulating transparent latches|
|US7406406 *||Dec 7, 2004||Jul 29, 2008||Bull Hn Information Systems Inc.||Instructions to load and store containing words in a computer system emulator with host word size larger than that of emulated machine|
|US7689403 *||Apr 17, 2008||Mar 30, 2010||Bull Hn||Instructions providing new functionality for utilization by a target system program of host system container words during computer system emulation with host word size larger than that of the emulated machine|
|US7962296 *||Sep 12, 2008||Jun 14, 2011||Drone Technology Pte Ltd||Module for data acquisition and control in a sensor/control network|
|US8001360 *||Jan 16, 2004||Aug 16, 2011||Microunity Systems Engineering, Inc.||Method and software for partitioned group element selection operation|
|US8074008 *||Sep 10, 2009||Dec 6, 2011||Broadcom Corporation||Motherboard with video data processing card capability|
|US8145470||Jun 27, 2007||Mar 27, 2012||Microsoft Corporation||Accelerated access device emulator access scheme in a hypervisor environment with child and root partitions|
|US20030036893 *||Mar 28, 2002||Feb 20, 2003||Chen Liang T.||Method and apparatus for simulating transparent latches|
|US20040153632 *||Jan 16, 2004||Aug 5, 2004||Microunity Systems Engineering, Inc.||Method and software for partitioned group element selection operation|
|US20040205323 *||Jan 15, 2004||Oct 14, 2004||Microunity Systems Engineering, Inc.||Programmable processor and method for partitioned group element selection operation|
|US20060155524 *||Dec 7, 2004||Jul 13, 2006||Guethner Russell W||Instructions to load and store containing words in a computer system emulator with host word size larger than that of the emulated machine|
|US20080208562 *||Apr 17, 2008||Aug 28, 2008||Bull Hn Information Systems Inc.||Instructions providing new functionality for utilization by a target system program of host system container words during computer system emulation with host word size larger than that of the emulated machine|
|US20090006074 *||Jun 27, 2007||Jan 1, 2009||Microsoft Corporation||Accelerated access to device emulators in a hypervisor environment|
|US20090271131 *||Oct 29, 2009||Drone Technology Pte Ltd||Module for data acquisition and control in a sensor/control network|
|US20100005210 *||Sep 10, 2009||Jan 7, 2010||Broadcom Corporation||Motherboard With Video Data Processing Card Capability|
|WO1991020041A1 *||Jun 10, 1991||Dec 26, 1991||Supercomputer Systems Ltd||Multiple request toggling priority arbitration system|
|U.S. Classification||703/26, 712/E09.006, 712/E09.037|
|International Classification||G06F9/318, G06F9/22|
|Cooperative Classification||G06F9/226, G06F9/3017|
|European Classification||G06F9/30U, G06F9/22F|
|Jun 26, 1984||CC||Certificate of correction|
|Dec 22, 1987||AS||Assignment|
Owner name: TORONTO-DOMINION BANK, THE
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR, INC., (SEE RECORD FOR REMAINING GRANTORS);REEL/FRAME:004829/0701
Effective date: 19871216
Owner name: TORONTO-DOMINION BANK, THE,STATELESS
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR, INC., (SEE RECORD FOR REMAINING GRANTORS);REEL/FRAME:004829/0701
Effective date: 19871216
|Nov 3, 1989||AS||Assignment|
Owner name: TRACOR AEROSPACE, INC., A CORP. OF TX.
Free format text: MERGER;ASSIGNORS:TRACOR AEROSPACE AUSTIN, INC. (CHANGED TO);TRACOR AEROSPACE, INC., A CORP. OF TX. (AND);TRACOR RADCON, INC., A CORP.OF TX. (MERGED INTO);AND OTHERS;REEL/FRAME:005224/0668
Effective date: 19861210
|Nov 13, 1989||AS||Assignment|
Owner name: TRACOR AEROSPACE, INC., A CORP. OF TX.
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST. EFFECTIVE OCTOBER 24, 1989;ASSIGNOR:TRACOR, INC., A CORP. OF DE.;REEL/FRAME:005224/0687
Effective date: 19891027
|Dec 13, 1989||AS||Assignment|
Owner name: BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIA
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR INC.;REEL/FRAME:005217/0224
Effective date: 19880801
Owner name: BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIA
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR, INC.;REEL/FRAME:005217/0247
Effective date: 19880801
Owner name: BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIA
Free format text: SECURITY INTEREST;ASSIGNORS:TORONTO-DOMINION BANK;TRACOR, INC.;REEL/FRAME:005224/0276
Effective date: 19880801
Owner name: BANK OF AMERICA AS AGENT
Free format text: SECURITY INTEREST;ASSIGNOR:TORONTO-DOMINION BANK, THE;REEL/FRAME:005197/0122
Effective date: 19880801
Owner name: BANK OF AMERICA NATIONAL TRUST AND SAVING ASSOCIAT
Free format text: SECURITY INTEREST;ASSIGNOR:TORONTO DOMINION BANK, THE,;REEL/FRAME:005284/0163
Effective date: 19880801
Owner name: TORONTO-DOMINION BANK, THE
Free format text: SECURITY INTEREST;ASSIGNORS:TRACOR, INC.;LITTLEFUSE, INC.;TRACOR AEROSPACE, INC.;AND OTHERS;REEL/FRAME:005234/0127
Effective date: 19880801
|Dec 19, 1989||AS||Assignment|
Owner name: BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIA
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR HOLDINGS, INC., TRACOR, INC., AND OTHERS INDICATED ON SCHEDULE SA;REEL/FRAME:005317/0726
Effective date: 19891030
|Dec 27, 1991||AS||Assignment|
Owner name: TRACOR, INC.
Free format text: RELEASED BY SECURED PARTY;ASSIGNOR:BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIATION AS COLLATERAL AGENT;REEL/FRAME:005957/0562
Effective date: 19911220
Owner name: TRACOR AEROSPACE, INC. A DE CORPORATION, TEXAS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNOR:OTC TRACOR AEROSPACE, INC., A TX CORPORATION;REEL/FRAME:005955/0330
Effective date: 19911220
Owner name: CONTINENTAL BANK N.A.
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR AEROSPACE, INC.;REEL/FRAME:005955/0268
Effective date: 19911227
Owner name: TRACOR, INC.
Free format text: RELEASED BY SECURED PARTY;ASSIGNOR:BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIATION;REEL/FRAME:005953/0942
Effective date: 19911227
Owner name: TRACOR, INC.
Free format text: RELEASED BY SECURED PARTY;ASSIGNOR:BANK OF AMERICA NATIONAL TRUST AND SAVINGS ASSOCIATION AS COLLATERAL AGENT;REEL/FRAME:005957/0542
Effective date: 19911227
Owner name: OTC TRACOR AEROSPACE, INC.
Free format text: CHANGE OF NAME;ASSIGNOR:TRACOR AEROSPACE, INC.;REEL/FRAME:005955/0250
Effective date: 19911122
|Sep 2, 1993||AS||Assignment|
Owner name: TRACOR AEROSPACE, INC., TEXAS
Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:CONTINENTAL BANK N.A.;REEL/FRAME:006683/0028
Effective date: 19930823
Owner name: BANKERS TRUST COMPANY, NEW YORK
Free format text: SECURITY INTEREST;ASSIGNOR:TRACOR AEROSPACE, INC.;REEL/FRAME:006674/0934
Effective date: 19930825
|May 22, 1996||AS||Assignment|
Owner name: BANKERS TRUST COMPANY, NEW YORK
Free format text: ASSIGNMENT OF SECURITY;ASSIGNOR:TRACOR AEROSPACE, INC.;REEL/FRAME:007969/0164
Effective date: 19960222
|Mar 20, 1997||AS||Assignment|
Owner name: TRACOR AEROSPACE, INC., TEXAS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BANKERS TRUST COMPANY;REEL/FRAME:008401/0898
Effective date: 19970314